Nothing
R/bma.R
In BayesianMediationA: Bayesian Mediation Analysis
Defines functions
summary.bma.bx.cy
bma.bx.cy
data_org
Documented in
bma.bx.cy
data_org
summary.bma.bx.cy
data_org<-function(pred, m, y, refy = rep(NA, ncol(data.frame(y))),
predref = rep(NA, ncol(data.frame(pred))), fpy = NULL,
deltap = rep(0.001,ncol(data.frame(pred))), fmy = NULL,
deltam = rep(0.001,ncol(data.frame(m))), fpm = NULL,
mref = rep(NA, ncol(data.frame(m))),cova = NULL,
mcov = NULL, mclist=NULL) #mcov is the data frame with all covariates for TVs, mind is the indicator for covariates
#if mclist is null but not mcov, mcov is applied to all tvs.
{ns.dev<-function (x, df = NULL, knots = NULL, qnots=NULL,intercept = FALSE, Boundary.knots = range(x),derivs1=0)
{
nx <- names(x)
x <- as.vector(x)
nax <- is.na(x)
if (nas <- any(nax))
x <- x[!nax]
if (!missing(Boundary.knots)) {
Boundary.knots <- sort(Boundary.knots)
outside <- (ol <- x < Boundary.knots[1L]) | (or <- x >
Boundary.knots[2L])
}
else outside <- FALSE
if (!is.null(df) && is.null(knots)) {
nIknots <- df - 1L - intercept
if (nIknots < 0L) {
nIknots <- 0L
warning(gettextf("'df' was too small; have used %d",
1L + intercept), domain = NA)
}
knots <- if (nIknots > 0L) {
knots <- seq.int(0, 1, length.out = nIknots + 2L)[-c(1L,
nIknots + 2L)]
stats::quantile(x[!outside], knots)
}
}
else {if(is.null(df) && is.null(knots) && !is.null(qnots))
knots<-quantile(x[!outside], qnots)
nIknots <- length(knots)}
Aknots <- sort(c(rep(Boundary.knots, 4L), knots))
if (any(outside)) {
basis <- array(0, c(length(x), nIknots + 4L))
if (any(ol)) {
k.pivot <- Boundary.knots[1L]
xl <- cbind(1, x[ol] - k.pivot)
tt <- splineDesign(Aknots, rep(k.pivot, 2L), 4, c(0,
1),derivs=rep(derivs1,2L))
basis[ol, ] <- xl %*% tt
}
if (any(or)) {
k.pivot <- Boundary.knots[2L]
xr <- cbind(1, x[or] - k.pivot)
tt <- splineDesign(Aknots, rep(k.pivot, 2L), 4, c(0,1),derivs=rep(derivs1,2L))
basis[or, ] <- xr %*% tt
}
if (any(inside <- !outside))
basis[inside, ] <- splineDesign(Aknots, x[inside],
4,derivs=rep(derivs1,length(x[inside])))
}
else basis <- splineDesign(Aknots, x, 4,derivs=rep(derivs1,length(x)))
const <- splineDesign(Aknots, Boundary.knots, 4, c(2, 2),derivs=rep(derivs1,length(Boundary.knots)))
if (!intercept) {
const <- const[, -1, drop = FALSE]
basis <- basis[, -1, drop = FALSE]
}
qr.const <- qr(t(const))
basis <- as.matrix((t(qr.qty(qr.const, t(basis))))[, -(1L:2L),
drop = FALSE])
n.col <- ncol(basis)
if (nas) {
nmat <- matrix(NA, length(nax), n.col)
nmat[!nax, ] <- basis
basis <- nmat
}
dimnames(basis) <- list(nx, 1L:n.col)
a <- list(degree = 3L, knots = if (is.null(knots)) numeric() else knots,
Boundary.knots = Boundary.knots, intercept = intercept)
attributes(basis) <- c(attributes(basis), a)
class(basis) <- c("ns", "basis", "matrix")
basis
}
bs.dev<-function (x, df = NULL, knots = NULL, degree = 3, intercept = FALSE,
Boundary.knots = range(x),derivs1=0)
{
nx <- names(x)
x <- as.vector(x)
nax <- is.na(x)
if (nas <- any(nax))
x <- x[!nax]
if (!missing(Boundary.knots)) {
Boundary.knots <- sort(Boundary.knots)
outside <- (ol <- x < Boundary.knots[1L]) | (or <- x >
Boundary.knots[2L])
}
else outside <- FALSE
ord <- 1L + (degree <- as.integer(degree))
if (ord <= 1)
stop("'degree' must be integer >= 1")
if (!is.null(df) && is.null(knots)) {
nIknots <- df - ord + (1L - intercept)
if (nIknots < 0L) {
nIknots <- 0L
warning(gettextf("'df' was too small; have used %d",
ord - (1L - intercept)), domain = NA)
}
knots <- if (nIknots > 0L) {
knots <- seq.int(from = 0, to = 1, length.out = nIknots +
2L)[-c(1L, nIknots + 2L)]
stats::quantile(x[!outside], knots)
}
}
Aknots <- sort(c(rep(Boundary.knots, ord), knots))
if (any(outside)) {
warning("some 'x' values beyond boundary knots may cause ill-conditioned bases")
derivs <- 0:degree
scalef <- gamma(1L:ord)
basis <- array(0, c(length(x), length(Aknots) - degree -
1L))
if (any(ol)) {
k.pivot <- Boundary.knots[1L]
xl <- cbind(1, outer(x[ol] - k.pivot, 1L:degree,
"^"))
tt <- splineDesign(Aknots, rep(k.pivot, ord), ord,
derivs+derivs1)
basis[ol, ] <- xl %*% (tt/scalef)
}
if (any(or)) {
k.pivot <- Boundary.knots[2L]
xr <- cbind(1, outer(x[or] - k.pivot, 1L:degree,
"^"))
tt <- splineDesign(Aknots, rep(k.pivot, ord), ord,
derivs+derivs1)
basis[or, ] <- xr %*% (tt/scalef)
}
if (any(inside <- !outside))
basis[inside, ] <- splineDesign(Aknots, x[inside],
ord,derivs=rep(derivs1,length(x[inside])))
}
else basis <- splineDesign(Aknots, x, ord, derivs=rep(derivs1,length(x)))
if (!intercept)
basis <- basis[, -1L, drop = FALSE]
n.col <- ncol(basis)
if (nas) {
nmat <- matrix(NA, length(nax), n.col)
nmat[!nax, ] <- basis
basis <- nmat
}
dimnames(basis) <- list(nx, 1L:n.col)
a <- list(degree = degree, knots = if (is.null(knots)) numeric(0L) else knots,
Boundary.knots = Boundary.knots, intercept = intercept)
attributes(basis) <- c(attributes(basis), a)
class(basis) <- c("bs", "basis", "matrix")
basis
}
x2fx<-function(x,func) #x is the list of original numerical vector, func is a vector of character functions.
{ # eg. func <- c("x","x+1","x+2","x+3","log(x)")
func.list <- list()
#test.data <- matrix(data=rep(x,length(func)),length(x),length(func))
#test.data <- data.frame(test.data)
result<-NULL
for(i in 1:length(func)){
func.list[[i]] <- function(x){}
body(func.list[[i]]) <- parse(text=func[i])
}
#result <- mapply(do.call,func.list,lapply(test.data,list))
col_fun<-NULL
z<-1
for (i in 1:length(func.list))
{res<-as.matrix(func.list[[i]](x))
result<-cbind(result,res)
col_fun<-cbind(col_fun,c(z,z+ncol(res)-1))
z<-z+ncol(res)}
list(values=as.matrix(result),col_fun=as.matrix(col_fun))
}
x2fdx<-function(x,func) #x is the list of original numerical vector, func is a vector of character functions.
{ fdx<-NULL # eg. func <- c("x","x+1","x+2","x+3","log(x)")
for(i in 1:length(func)){
if (length(grep("ifelse",func[i]))>0)
{str<-unlist(strsplit(func[i],","))
fun1<-D(parse(text=str[2]), "x")
fun2<-D(parse(text=unlist(strsplit(str[3],")"))),"x")
x1<-eval(fun1)
x2<-eval(fun2)
if(length(x1)==1)
x1<-rep(x1,length(x))
if(length(x2)==1)
x2<-rep(x2,length(x))
fun3<-paste(str[1],"x1,x2)",sep=",")
fdx<-cbind(fdx,eval(parse(text=fun3)))
}
else if(length(grep("ns",func[i]))>0)
{temp<-paste("ns.dev",substring(func[i],3,nchar(func[i])-1),",derivs1=1)",sep="")
fdx<-cbind(fdx,eval(parse(text=temp)))}
else if(length(grep("bs",func[i]))>0)
{temp<-paste("bs.dev",substring(func[i],3,nchar(func[i])-1),",derivs1=1)",sep="")
fdx<-cbind(fdx,eval(parse(text=temp)))}
else{
dx2x <- D(parse(text=func[i]), "x")
temp<-eval(dx2x)
if(length(temp)==1)
fdx<-cbind(fdx,rep(temp,length(x)))
else fdx<-cbind(fdx,temp)}
}
as.matrix(fdx)
}
bin_cat <- function(M2, M1, cat1, catref = 1)
#turn a categorical variable to binary dummy variables
#M2 is the original data frame
#cat1 is the column number of the categorical variable in M2
#catref is the reference group
{a <- factor(M2[, cat1])
b <- sort(unique(a[a != catref]))
d<-NULL
e<-rep(1,nrow(M2))
for(i in 1:length(b))
{d=cbind(d,ifelse(a==b[i],1,0))
e=ifelse(a==b[i],i+1,e)
}
d[is.na(M2[,cat1]),]=NA
e[is.na(M2[,cat1])]=NA
M2[,cat1]=e
xnames=colnames(M1)
M1=cbind(M1,d)
colnames(M1)=c(xnames,paste(colnames(M2)[cat1],b,sep="."))
list(M1=M1,M2=M2,cat=c(ncol(M1)-ncol(d)+1,ncol(M1)))
}
order_char<-function(char1,char2) #find the position of char2 in char1
{a<-1:length(char1)
b<-NULL
for (i in 1:length(char2))
b<-c(b,a[char1==char2[i]])
b
}
###start the main code
#clean up the outcomes:y_type=type of outcomes;
#y_type=2:binary, 3:category, 1:continuous, 4:time-to-event
#consider only 1 outcome for now
#consider only complete case
data.temp=cbind(pred,m,y)
if (!is.null(mcov) & !is.null(mclist))
data.temp=cbind(pred,m,y,mcov)
if(!is.null(cova))
data.temp=cbind(data.temp,cova)
choose.temp=complete.cases(data.temp)
if(ncol(data.frame(pred))==1)
pred=pred[choose.temp]
else
pred=pred[choose.temp,]
if(ncol(data.frame(y))==1)
y=y[choose.temp]
else
y=y[choose.temp,]
if(ncol(data.frame(m))==1)
m=m[choose.temp]
else
m=m[choose.temp,]
if(!is.null(cova)){
if(ncol(data.frame(cova))==1)
cova=cova[choose.temp]
else
cova=cova[choose.temp,]}
if(!is.null(mcov)){
if(ncol(data.frame(mcov))==1)
mcov=mcov[choose.temp]
else
mcov=mcov[choose.temp,]}
if (!is(y,"Surv"))
{if (nlevels(droplevels(as.factor(y))) == 2) {#binary
y_type <- 2
if (!is.na(refy))
y <- ifelse(y == refy, 0, 1)
else {
refy <- levels(droplevels(as.factor(y)))[1]
y <- ifelse(as.factor(y) == refy,0, 1)
}
}
else if (is.character(y) | is.factor(y)) {#categorical
y_type <- 3
y <- droplevels(y)
if (is.na(refy))
refy <- levels(as.factor(y))[1]
a <- factor(y)
b <- sort(unique(a[a != refy]))
e <- rep(1,nrow(as.matrix(y)))
for(j in 1:length(b))
e=ifelse(a==b[j],j+1,e)
e[is.na(y)]=NA
y=factor(e)
}
else #continuous
y_type = 1}
else
{y_type<-4
y=cbind(y[,1],y[,2])}
#clean up m and pred
mnames <- colnames(m)
if (!is.null(cova)) {
if (is.null(colnames(cova)))
cova_names = paste("cova",1:ncol(data.frame(cova)))
else
cova_names = colnames(cova)
cova=data.frame(cova)
colnames(cova)=cova_names
}
if (!is.null(mcov)) {
if (length(grep("for.m", names(mcov))) == 0)
mcov_names = colnames(mcov)
else mcov_names = colnames(mcov[[1]])
mcov=data.frame(mcov)
}
pred_names = names(pred)
##prepare for the predictor(s)
pred1 <- data.frame(pred) #original format
pred1.0 <- NULL #original format with binarized categorical predictor
pred1.0_names <-NULL
pred2 <- NULL #all transformed
pred2_names = NULL
pred3 <- NULL #transformed continuous predictors with pred+delta(pred)
pred3_names = NULL
pred.cont.der <- NULL #derivative of the transformation function for cont predictors
if (is.null(pred_names))
pred_names = paste("pred",1:ncol(pred1),sep='')
colnames(pred1) = pred_names
binpred = NULL #binary predictor in pred2
catpred = NULL #categorical predictor in pred2, each row is for one categorical predictor
contpred = NULL #continuos predictor in pred2, each row is for a continuous predictor
binpred1 = NULL #binary predictor in pred1
catpred1 = NULL #categorical predictor in pred1
contpred1 = NULL #continuous predictor in pred1
binpred1.0 = NULL #binary predictor in pred1.0
catpred1.0 = NULL #categorical predictor in pred1.0
contpred1.0 = NULL #continuous predictor in pred1.0
contpred3 = NULL #index for pred3 and pred.cont.dev
npred = ncol(pred1)
n1=nrow(pred1)
for (i in 1:npred)
if (nlevels(droplevels(as.factor(pred1[,i]))) == 2) { #binary predictor
if (!is.na(predref[i]))
{pred2 <- cbind(pred2,ifelse(pred1[, i] == predref[i],0, 1))
pred1.0 <- cbind(pred1.0, ifelse(pred1[, i] == predref[i],0, 1))
#pred3 <- cbind(pred3,as.factor(ifelse(pred1[, i] == predref[i],0, 1)))
pred1[,i] <- ifelse(pred1[, i] == predref[i],0, 1)}
else {
temp.pred <- as.factor(pred1[, i])
pred2 <- cbind(pred2,ifelse(temp.pred == levels(droplevels(temp.pred))[1], 0, 1))
pred1.0 <- cbind(pred1.0,ifelse(temp.pred == levels(droplevels(temp.pred))[1], 0, 1))
#pred3 <- cbind(pred3,as.factor(ifelse(temp.pred == levels(droplevels(temp.pred))[1], 0, 1)))
pred1[,i] <- ifelse(temp.pred == levels(droplevels(temp.pred))[1], 0, 1)
}
binpred1 = c(binpred1, i)
pred2_names=c(pred2_names,pred_names[i])
pred1.0_names=c(pred1.0_names,pred_names[i])
binpred = c(binpred,ncol(pred2))
binpred1.0 = c(binpred1.0,ncol(pred1.0))
}
else if (is.character(pred1[, i]) | is.factor(pred1[, i])) { #category predictor
pred1[, i] = droplevels(pred1[, i])
if(!is.null(pred2))
colnames(pred2)=pred2_names
if(!is.null(pred1.0))
colnames(pred1.0)=pred1.0_names
# catn = catn + 1
if (!is.na(predref[i]))
{pred.temp1 <- bin_cat(pred1, pred2, i, predref[i])
pred.temp2 <- bin_cat(pred1, pred1.0, i, predref[i])}
else
{pred.temp1 <- bin_cat(pred1, pred2, i, levels(as.factor(pred1[,i]))[1])
pred.temp2 <- bin_cat(pred1, pred1.0, i, levels(as.factor(pred1[,i]))[1])}
pred2 = pred.temp1$M1
pred1.0 = pred.temp2$M1
pred1 = pred.temp1$M2
#pred3 = cbind(pred3,pred.temp1$M1[,pred.temp1$cat[1]:pred.temp1$cat[2]])
catpred1 = c(catpred1,i)
catpred = rbind(catpred,pred.temp1$cat)
catpred1.0 = rbind(catpred1.0,pred.temp2$cat)
pred2_names = colnames(pred.temp1$M1)
pred1.0_names = colnames(pred.temp2$M1)
}
else #consider the transformation of continuous x
{contpred1 = c(contpred1, i)
pred1.0=cbind(pred1.0,pred1[,i])
pred1.0_names<-c(pred1.0_names,pred_names[i])
contpred1.0= c(contpred1.0, ncol(pred1.0))
if(!(i %in% fpy[[1]])) #fpy has the transformation functions for pred to y, [[1]] list all cont pred to be transformed
{pred2<-cbind(pred2,pred1[,i])
pred.cont.der<-cbind(pred.cont.der,rep(1,n1))
contpred3 = rbind(contpred3,c(ncol(pred.cont.der),ncol(pred.cont.der)))
pred3 = cbind(pred3,pred1[,i]+deltap[i]) #deltap is the changing amount for predictors
contpred = rbind(contpred,c(ncol(pred2),ncol(pred2)))
pred2_names=c(pred2_names,pred_names[i])
pred3_names=c(pred3_names,pred_names[i])
}
else if (i %in% fpy[[1]])
{p=match(i,fpy[[1]])
temp.pred=x2fx(pred1[,i],fpy[[1+p]])$values
pred2<-cbind(pred2,temp.pred)
pred.cont.der<-cbind(pred.cont.der,x2fdx(pred1[,i],fpy[[1+p]]))
if(is.null(pred3))
contpred3 = rbind(contpred3,c(1,ncol(pred.cont.der)))
else
contpred3 = rbind(contpred3,c(ncol(pred3)+1,ncol(pred.cont.der)))
pred3<-cbind(pred3,x2fx(pred1[,i]+deltap[i],fpy[[1+p]])$values)
l=length(fpy[[1+p]])
contpred=rbind(contpred,c(ncol(pred2)-ncol(temp.pred)+1,ncol(pred2)))
pred2_names<-c(pred2_names,paste(pred_names[i],1:l,sep="."))
pred3_names=c(pred3_names,paste(pred_names[i],1:l,sep="."))
}
}
colnames(pred1.0)=pred1.0_names
colnames(pred2)=pred2_names
if(!is.null(pred3))
{colnames(pred3)=pred3_names
colnames(pred.cont.der)=pred3_names}
## prepare mediators for y
m1 <- data.frame(m) #original format
m2 <- NULL #all transformed
m2_names = NULL
m3 <- NULL #transformed continuous mediators with mediator+delta(med)
m3_names = NULL
m.cont.der <- NULL #derivative of the transformation function for cont mediators
if (is.null(mnames))
mnames = paste("m",1:ncol(m1),sep='')
colnames(m1) = mnames
binm = NULL
catm = NULL
contm = NULL
binm1 = NULL
catm1 = NULL
contm1 = NULL
contm3 = NULL #index for m3 and m.cont.dev
nm = ncol(m1)
n2=nrow(m1)
for (i in 1:nm)
if (nlevels(droplevels(as.factor(m1[,i]))) == 2) { #binary mediator
if (!is.na(mref[i]))
{m2 <- cbind(m2,ifelse(m1[, i] == mref[i],0, 1))
#m3 <- cbind(m3,as.factor(ifelse(m1[, i] == mref[i],0, 1)))
m1[,i] <- ifelse(m1[, i] == mref[i],0, 1)}
else {
temp.m <- as.factor(m1[, i])
m2 <- cbind(m2,ifelse(temp.m == levels(droplevels(temp.m))[1], 0, 1))
#m3 <- cbind(m3,as.factor(ifelse(temp.m == levels(droplevels(temp.m))[1], 0, 1)))
m1[,i] <- ifelse(temp.m == levels(droplevels(temp.m))[1], 0, 1)
}
binm1 = c(binm1, i)
m2_names=c(m2_names,mnames[i])
colnames(m2)=m2_names
binm = c(binm,ncol(m2))
}
else if (is.character(m1[, i]) | is.factor(m1[, i])) { #category mediator
m1[, i] = droplevels(as.factor(m1[, i]))
if (!is.na(mref[i]))
m.temp1 <- bin_cat(m1, m2, i, mref[i])
else
m.temp1 <- bin_cat(m1, m2, i, levels(as.factor(m1[,i]))[1])
m2 = m.temp1$M1
m1 = m.temp1$M2
#m3 = cbind(m3,m.temp1$M1[,m.temp1$cat[1]:m.temp1$cat[2]])
catm1 = c(catm1,i)
catm = rbind(catm,m.temp1$cat)
m2_names = c(m2_names, colnames(m.temp1$M1)[m.temp1$cat[1]:m.temp1$cat[2]])
colnames(m2)=m2_names
}
else #consider the transformation of continuous m
{contm1 = c(contm1, i)
if(!(i %in% fmy[[1]])) #fmy has the transformation functions for m to y, [[1]] list all cont mediators in m to be transformed
{m2<-cbind(m2,m1[,i])
m.cont.der<-cbind(m.cont.der,rep(1,n1))
contm3 = rbind(contm3,c(ncol(m.cont.der),ncol(m.cont.der)))
m3 = cbind(m3,m1[,i]+deltam[i]) #deltam is the changing amount for mediators
m2_names=c(m2_names,mnames[i])
colnames(m2)=m2_names
m3_names=c(m3_names,mnames[i])
contm=rbind(contm,c(ncol(m2),ncol(m2)))
}
else if (i %in% fmy[[1]])
{p=match(i,fmy[[1]])
temp.m=x2fx(m1[,i],fmy[[1+p]])$values
m2<-cbind(m2,temp.m)
m.cont.der<-cbind(m.cont.der,x2fdx(as.matrix(m)[,i],fmy[[1+p]]))
if(is.null(m3))
contm3 = rbind(contm3,c(1,ncol(m.cont.der)))
else
contm3 = rbind(contm3,c(ncol(m3)+1,ncol(m.cont.der)))
m3<-cbind(m3,x2fx(m1[,i]+deltap[i],fmy[[1+p]])$values)
contm = rbind(contm,c(ncol(m2)-ncol(temp.m)+1,ncol(m2)))
l=length(fmy[[1+p]])
m2_names<-c(m2_names,paste(mnames[i],1:l,sep="."))
colnames(m2)=m2_names
m3_names=c(m3_names,paste(mnames[i],1:l,sep="."))
}
}
colnames(m2)=m2_names
if(!is.null(m3))
{colnames(m3)=m3_names
colnames(m.cont.der)=m3_names}
## prepare predictors for mediators
pm=rep(0,n1) #the first row are all 0s
fpm.2=fpm #the transformed predictors in pm
binp=NULL
catp=NULL
contp=NULL
j=1
names.pm=("zero")
if(!is.null(binpred))
for (i in binpred)
{pm=cbind(pm,pred2[,i])
j=j+1
binp=c(binp,j)
names.pm=c(names.pm,pred_names[i])}
if(!is.null(catpred))
for (i in 1:nrow(catpred))
{pm=cbind(pm,pred2[,catpred[i,1]:catpred[i,2]])
catp=rbind(catp,c(j+1,j+1+catpred[i,2]-catpred[i,1]))
j=j+1+catpred[i,2]-catpred[i,1]
names.pm=c(names.pm,colnames(pred2)[catpred[i,1]:catpred[i,2]])
}
if(!is.null(contpred1))
for (i in contpred1)
{pm=cbind(pm,pred1[,i])
j=j+1
contp=rbind(contp,rep(j,3))
names.pm=c(names.pm,pred_names[i])}
pm.der=matrix(1,n1,ncol(pm))
pm.idx=as.list(rep(1,ncol(m1))) #the predictors in pm to predict the ith mediator
for (i in 1:ncol(m1))
pm.idx[[i]]=2:ncol(pm)
if(!is.null(fpm))
{k=unique(fpm[[1]][,2]) #the first column is for the mediators,
#the second column the continuous predictors to be transformed
for (l in k){
temp<-(2:length(fpm))[fpm[[1]][,2]==l]
allfun=fpm[[temp[1]]]
if (length(temp)>1)
for(i in 2:length(temp))
allfun<-c(allfun,fpm[[temp[i]]])
unifun<-unique(allfun)
unifun1<-unifun[unifun!="x"]
unifun2<-c("x",unifun1)
d_d<-x2fx(pred1[,l],unifun1)
d.der<-x2fdx(pred1[,l],unifun1)
d<-as.matrix(d_d$values)
names.pm<-c(names.pm,paste(pred_names[l],1:ncol(d),sep="."))
pm.der<-cbind(pm.der,d.der)
place=match(l,contpred1)
contp[place,2]=j+1
contp[place,3]=j+ncol(d)
pm<-cbind(pm,d)
for(i in temp)
{ttemp<-order_char(unifun1,fpm[[i]])
fpm.2[[i]]=j+ttemp #what does this do?
pm.indx[[fpm[[1]][i-1,1]]]=c(pm.indx[[fpm[[1]][i-1,1]]],j+ttemp)
if(length(order_char('x',fpm[[i]]))==0)
pm.indx[[fpm[[1]][i-1,1]]]= (pm.indx[[fpm[[1]][i-1,1]]])[pm.indx[[fpm[[1]][i-1,1]]]!=l]
}
j=j+ncol(d)
}}
colnames(pm)=names.pm
colnames(pm.der)=names.pm
pm.ind=matrix(1,length(pm.idx),max(sapply(pm.idx,length)))
for (i in 1:nrow(pm.ind))
pm.ind[i,1:length(pm.idx[[i]])]=pm.idx[[i]]
p2=ifelse(is.null(binm1),0,length(binm1))
p3=ifelse(is.null(catm1),0,length(catm1))
p1=ifelse(is.null(contm1),0,length(contm1))
#prepare for covariates of mediators
if(is.null(mcov))
{mcov=data.frame(intercept=rep(1,nrow(m1)))
mind=matrix(T,p1+p2+p3,1)}
else
{mcov=data.frame(intercept=rep(1,nrow(m)),mcov)
mind=matrix(T,p1+p2+p3,ncol(mcov))
mcov_names=colnames(mcov)
if (!is.null(mclist))
{if (is.character(mclist[[1]]))
mclist[[1]]=match(mclist[[1]],mnames)
mcov=data.frame(mcov,no=rep(0,nrow(mcov))) #add a column of 0 in mcov
mind=matrix(rep(1:(ncol(mcov)-1),each=p1+p2+p3),p1+p2+p3,ncol(mcov)-1)
for (i in 1:length(mclist[[1]]))
{if(sum(is.na(mclist[[i+1]]))>=1)
temp=1
else if(is.character(mclist[[i+1]]))
temp=c(1,match(mclist[[i+1]],mcov_names))
else
temp=c(1,mclist[[i+1]]+1)
mind[mclist[[1]][i],(1:(ncol(mcov)-1))[-temp]]=ncol(mcov)
}}
} #use all covariates for all tv if mclist is NULL. Otherwise, the first item of mclist lists all tvs
#that are using different mcov, the following items gives the mcov for the tv in order. use NA is no mcov to be used
results = list(N=nrow(data.frame(y)), y_type=y_type, y=y, pred1=pred1, pred1.0=pred1.0,
pred2=pred2, pred3=pred3, cova=cova,
pred.cont.der=pred.cont.der, binpred2=binpred, catpred2=catpred,
contpred2=contpred, binpred1=binpred1, catpred1=catpred1,contpred1=contpred1,
contpred1.0=contpred1.0, binpred1.0=binpred1.0, catpred1.0=catpred1.0,
contpred3=contpred3, npred=npred,
m1=m1, m2=m2, m3=m3, m.cont.der=m.cont.der, binm2=binm, catm2=catm,
contm2=contm,binm1=binm1, catm1=catm1, contm1=contm1, contm3=contm3,
nm=nm, pm=pm, pm.der=pm.der, pm.idx=pm.idx, pm.ind=pm.ind, fpm.2=fpm.2,
binp=binp, catp=catp, contp=contp,p1=p1,p2=p2,p3=p3,mcov=mcov,mind=mind)
return(results)
}
bma.bx.cy<-function(pred, m, y, refy = rep(NA, ncol(data.frame(y))),
predref = rep(NA, ncol(data.frame(pred))), fpy = NULL,
deltap = rep(0.001,ncol(data.frame(pred))), fmy = NULL,
deltam = rep(0.001,ncol(data.frame(m))), fpm = NULL,
mref = rep(NA, ncol(data.frame(m))),cova = NULL,
mcov = NULL, mclist=NULL, inits=NULL,
n.chains = 1, n.iter = 1100,n.burnin=100,n.thin = 1,
mu=NULL, Omega=NULL, Omegac=NULL,muc=NULL,mucv=NULL, Omegacv=NULL,
mu0.1=NULL, Omega0.1=NULL, mu1.1=NULL, Omega1.1=NULL,
mu0.a=NULL, Omega0.a=NULL, mu1.a=NULL, Omega1.a=NULL,
mu0.b=NULL, Omega0.b=NULL, mu1.b=NULL, Omega1.b=NULL,
mu0.c=NULL, Omega0.c=NULL, mu1.c=NULL, Omega1.c=NULL,
preci=0.000001,tmax=Inf,multi=NULL,filename=NULL)
{ ### build the bugs model if it is not defined
jags_model <- function (contm=c('ncontm', 'ycontm'),binm=c('nbinm','ybinm'),
catm=c('ncatm','ycatm'),
cova=c('ncova','ycova'),
ytype=c('contc','binc','catc','survc','contnnc',
'binnc','catnc','survnc')) {
contm <- match.arg(contm)
binm <- match.arg(binm)
catm <- match.arg(catm)
cova <- match.arg(cova)
# raw script
script <-
"model {temp <- 1:nmc
for(i in 1:N){
$yfunc
$contm
$binm
$catm
}
$ypriors
$cntmprior
$bmprior
$cmprior
var4 ~ dgamma(1,0.1)
prec4 <-1/var4
}"
# define macros
macros <- list(list("$yfunc",
switch(ytype,
contc='mu_y[i] <- beta0 + inprod(c, x[i,]) + inprod(beta,M1[i,]) + inprod(eta,cova[i,])
y[i] ~ dnorm(mu_y[i],prec4)',
contnc='mu_y[i] <- beta0 + inprod(c, x[i,]) + inprod(beta,M1[i,])
y[i] ~ dnorm(mu_y[i],prec4)',
binc='logit(mu_y[i]) <- beta0 + inprod(c, x[i,]) + inprod(beta,M1[i,]) + inprod(eta,cova[i,])
y[i] ~ dbern(mu_y[i])',
binnc='logit(mu_y[i]) <- beta0 + inprod(c, x[i,]) + inprod(beta,M1[i,])
y[i] ~ dbern(mu_y[i])',
catc='mu_y1[i,1] <- 1
for (k in 2:caty)
{mu_y1[i,k] <- exp(beta0[k-1] + inprod(c[k-1,], x[i,]) + inprod(beta[k-1,],M1[i,]) + inprod(eta[k-1,],cova[i,]))}
sum_y[i] <- sum(mu_y1[i,1:caty])
for (l in 1:caty)
{mu_y[i,l] <- mu_y1[i,l]/sum_y[i]}
y[i] ~ dcat(mu_y[i,])',
catnc='mu_y1[i,1] <- 1
for (k in 2:caty)
{mu_y1[i,k] <- exp(beta0[k-1] + inprod(c[k-1,], x[i,]) + inprod(beta[k-1,],M1[i,]))}
sum_y[i] <- sum(mu_y1[i,1:caty])
for (l in 1:caty)
{mu_y[i,l] <- mu_y1[i,l]/sum_y[i]}
y[i]~dcat(mu_y[i,])',
survc= 'elinpred[i] <- exp(inprod(c, x[i,]) + inprod(beta,M1[i,]) + inprod(eta,cova[i,]))
base[i] <- lambda*r*pow(y[i,1], r-1)
loghaz[i] <- log(base[i]*elinpred[i])
phi[i] <- 100000-y[i,2]*loghaz[i]-log(exp(-lambda*pow(y[i,1],r)*elinpred[i])-exp(-lambda*pow(tmax,r)*elinpred[i])) +log(1-exp(-lambda*pow(tmax,r)*elinpred[i]))
zero[i] ~ dpois(phi[i])' ,
survnc= 'elinpred[i] <- exp(inprod(c, x[i,]) + inprod(beta,M1[i,]))
base[i] <- lambda*r*pow(y[i,1], r-1)
loghaz[i] <- log(base[i]*elinpred[i])
phi[i] <- 100000-y[i,2]*loghaz[i]-log(exp(-lambda*pow(y[i,1],r)*elinpred[i])-exp(-lambda*pow(tmax,r)*elinpred[i])) +log(1-exp(-lambda*pow(tmax,r)*elinpred[i]))
zero[i] ~ dpois(phi[i])')),
list("$contm",
switch(contm,
ycontm='for (j in 1:p1){
mu_M1[i,contm[j]] <- inprod(alpha0.a[j,mind[contm[j],]],mcov[i,temp[mind[contm[j],]]])+inprod(alpha1.a[j,1:c1],x1[i,])
M2[i,contm[j]] ~ dnorm(mu_M1[i,contm[j]],prec1[j])
for (k in contm1[j,1]:contm1[j,2]){
mu_M1_c[i,k] <- inprod(alpha0[k,mind[contm[j],]],mcov[i,mind[contm[j],]])+inprod(alpha1[k,1:c1],x1[i,])
M1[i,k] ~ dnorm(mu_M1_c[i,k],prec2[k])
}
}
',
ncontm='')),
list("$cntmprior",
switch(contm,
ycontm=' for(j in 1:P)
{alpha1[j,1:c1] ~ dmnorm(mu1.1[j,1:c1], Omega1.1[1:c1, 1:c1])
alpha0[j,1:nmc] ~ dmnorm(mu0.1[j,1:nmc], Omega0.1[1:nmc, 1:nmc])}
for (i in 1:P){
var2[i] ~ dgamma(1,0.1)
prec2[i] <- 1/var2[i]
}
for(j in 1:p1)
{alpha1.a[j,1:c1] ~ dmnorm(mu1.a[j,1:c1], Omega1.a[1:c1, 1:c1])
alpha0.a[j,1:nmc] ~ dmnorm(mu0.a[j,1:nmc], Omega0.a[1:nmc, 1:nmc])}
for (i in 1:p1){
var1[i] ~ dgamma(1,0.1)
prec1[i] <- 1/var1[i]
}',
ncontm='')),
list("$binm",
switch(binm,
ybinm=" for (k in 1:p2){
logit(mu_M1[i,binm[k]]) <- inprod(alpha0.b[k,mind[binm[k],]],mcov[i,mind[binm[k],]])+inprod(alpha1.b[k,1:c1],x1[i,])
M2[i,binm[k]] ~ dbern(mu_M1[i,binm[k]])
}",
nbinm='')),
list("$bmprior",
switch(binm,
nbinm='',
ybinm=' for(j in 1:p2)
{alpha1.b[j,1:c1] ~ dmnorm(mu1.b[j,1:c1], Omega1.b[1:c1, 1:c1])
alpha0.b[j,1:nmc] ~ dmnorm(mu0.b[j,1:nmc], Omega0.b[1:nmc, 1:nmc])
}')),
list("$catm",
switch(catm,
ycatm=" for (j in 1:p3){
mu_Mc[i,j,1] <- 1 #baseline is the 1st category
for (k in 2:cat2[j]){
mu_Mc[i,j,k] <- exp(inprod(alpha0.c[j,k-1,mind[catm[j],]],mcov[i,mind[catm[j],]])+inprod(alpha1.c[j,k-1,1:c1],x1[i,]))
}
sum_Mc[i,j] <- sum(mu_Mc[i,j,1:cat2[j]])
for (l in 1:cat2[j])
{mu_Mc0[i,j,l] <- mu_Mc[i,j,l]/sum_Mc[i,j]}
M2[i,catm[j]] ~ dcat(mu_Mc0[i,j,1:cat2[j]])
}",
ncatm='')),
list("$cmprior",
switch(catm,
ncatm='',
ycatm=' for (i in 1:p3){
for(j in 1:cat1)
{alpha1.c[i,j,1:c1] ~ dmnorm(mu1.c[j,1:c1], Omega1.c[1:c1, 1:c1])
alpha0.c[i,j,1:nmc] ~ dmnorm(mu0.c[j,1:nmc], Omega0.c[1:nmc, 1:nmc])}
}')),
list("$ypriors",
switch(ytype,
contc=' beta[1:P] ~ dmnorm(mu[1:P], Omega[1:P, 1:P])
beta0 ~ dnorm(0, 1.0E-6)
c[1:c2] ~ dmnorm(muc[1:c2], Omegac[1:c2,1:c2])
eta[1:cv1] ~ dmnorm(mucv[1:cv1], Omegacv[1:cv1,1:cv1])',
contnc=' beta[1:P] ~ dmnorm(mu[1:P], Omega[1:P, 1:P])
beta0 ~ dnorm(0, 1.0E-6)
c[1:c2] ~ dmnorm(muc[1:c2], Omegac[1:c2,1:c2])
',
binc=' beta[1:P] ~ dmnorm(mu[1:P], Omega[1:P, 1:P])
beta0 ~ dnorm(0, 1.0E-6)
c[1:c2] ~ dmnorm(muc[1:c2], Omegac[1:c2,1:c2])
eta[1:cv1] ~ dmnorm(mucv[1:cv1], Omegacv[1:cv1,1:cv1])',
binnc=' beta[1:P] ~ dmnorm(mu[1:P], Omega[1:P, 1:P])
beta0 ~ dnorm(0, 1.0E-6)
c[1:c2] ~ dmnorm(muc[1:c2], Omegac[1:c2,1:c2])
',
catc=' for(j in 1:(caty-1))
{beta[j,1:P] ~ dmnorm(mu[1:P], Omega[1:P, 1:P])
beta0[j] ~ dnorm(0, 1.0E-6)
c[j,1:c2] ~ dmnorm(muc[1:c2], Omegac[1:c2,1:c2])
eta[j,1:cv1] ~ dmnorm(mucv[1:cv1], Omegacv[1:cv1,1:cv1])}',
catnc=' for(j in 1:(caty-1))
{beta[j,1:P] ~ dmnorm(mu[1:P], Omega[1:P, 1:P])
beta0[j] ~ dnorm(0, 1.0E-6)
c[j,1:c2] ~ dmnorm(muc[1:c2], Omegac[1:c2,1:c2])}',
survc= ' beta[1:P] ~ dmnorm(mu[1:P], Omega[1:P, 1:P])
beta0 ~ dnorm(0, 1.0E-6)
c[1:c2] ~ dmnorm(muc[1:c2], Omegac[1:c2,1:c2])
r~duif(0,10) # dunif(0.5,1.5)
lambda~dgamma(1,0.01)
eta[1:cv1] ~ dmnorm(mucv[1:cv1], Omegacv[1:cv1,1:cv1])',
survnc= ' beta[1:P] ~ dmnorm(mu[1:P], Omega[1:P, 1:P])
beta0 ~ dnorm(0, 1.0E-6)
c[1:c2] ~ dmnorm(muc[1:c2], Omegac[1:c2,1:c2])
r ~ dunif(0,10) # dunif(1,1.2)
lambda ~ dgamma(1,0.01)')) # dunif(1.0E-8,3.0E-7)
)
# apply macros
for (m in seq(macros)) {
script <- gsub(macros[[m]][1], macros[[m]][2], script, fixed=TRUE)
}
script
}
### Incomplete Gamma function
Igamma<-function(z,u)pgamma(u,z)*gamma(z)
### Truncated Weibull distribution moments
weib.trunc.mean<-function(vec, right)vec[-1]*Igamma(1/vec[1]+1,(right/vec[-1])^vec[1])/(1-exp(-(right/vec[-1])^vec[1]))
matrix.prod<-function(m1)
{return(m1[1,]%*%t(m1[-1,]))}
expp<-function(v1,v2)
{v1^v2}
data0<- data_org(pred=pred, m=m, y=y, refy=refy, predref=predref, fpy=fpy,
deltap=deltap, fmy=fmy, deltam=deltam, fpm=fpm, mref=mref,
cova=cova,mcov = mcov, mclist=mclist)
y.type=data0$y_type #1 for continuous outcome, 4 for time-to-event, 2 for binary, 3 for categorical
N=data0$N
x=data0$pred2
c2=ncol(x)
x1=data0$pred1.0
c1=ncol(x1) #c1 is the number of predictors, k-class categorical variable are considered as k-1 predictors
y=data0$y
M1=data0$m2
M2=data0$m1
M3=data0$m3
contm=data0$contm1
contm1=data0$contm2
contm3=data0$contm3
p1=data0$p1
binm=data0$binm1
binm1=data0$binm2
p2=data0$p2
p3=data0$p3
if(p3>0){
cat1=max(data0$catm2[,2]-data0$catm2[,1]+1)
cat2=data0$catm2[,2]-data0$catm2[,1]+2
catm=data0$catm1
catm1=data0$catm2}
else{
cat1=NULL
cat2=NULL
catm=NULL
catm1=NULL
}
P=ncol(data0$m2)
cova=data0$cova
mcov=data0$mcov
mind=data0$mind
nmc=ncol(mcov)
# pm=data0$pm
# pm.ind=data1$pm.ind
if(is.null(mu))
mu=rep(0,P)
if(is.null(Omega))
Omega=diag(preci,P)
if(p1>0){
if(is.null(mu0.1))
mu0.1=matrix(0,P,nmc)
if(is.null(Omega0.1))
Omega0.1=diag(preci,nmc)
if(is.null(mu1.1))
mu1.1=matrix(0,P,c1)
if(is.null(Omega1.1))
Omega1.1=diag(preci,c1)
if(is.null(mu0.a))
mu0.a=matrix(0,p1,nmc)
if(is.null(Omega0.a))
Omega0.a=diag(preci,nmc)
if(is.null(mu1.a))
mu1.a=matrix(0,p1,c1)
if(is.null(Omega1.a))
Omega1.a=diag(preci,c1)
}
if(p2>0){
if(is.null(mu0.b))
mu0.b=matrix(0,p2,nmc)
if(is.null(Omega0.b))
Omega0.b=diag(preci,nmc)
if(is.null(mu1.b))
mu1.b=matrix(0,p2,c1)
if(is.null(Omega1.b))
Omega1.b=diag(preci,c1)
}
if(p3>0){
if(is.null(mu0.c))
mu0.c=matrix(0,cat1,nmc)
if(is.null(Omega0.c))
Omega0.c=diag(preci,nmc)
if(is.null(mu1.c))
mu1.c=matrix(0,cat1,c1)
if(is.null(Omega1.c))
Omega1.c=diag(preci,c1)
}
if(is.null(muc))
muc=rep(0,c2)
if(is.null(Omegac))
Omegac=diag(preci,c2)
# data0.1<- list (N=N,x=x,y=y,M1=M1,M2=M2,contm=contm,contm1=contm1,p1=p1,
# binm=binm,p2=p2,cat1=cat1,cat2=cat2,catm=catm,p3=p3,P=P,
# mu=mu,Omega=Omega,mu0.1=mu0.1,mu1.1=mu1.1,Omega0.1=Omega0.1,Omega1.1=Omega1.1,
# mu0.a=mu0.a,mu1.a=mu1.a,Omega0.a=Omega0.a,Omega1.a=Omega1.a,
# mu0.b=mu0.b,mu1.b=mu1.b,Omega0.b=Omega0.b,Omega1.b=Omega1.b,
# mu0.c=mu0.c,mu1.c=mu1.c,Omega0.c=Omega0.c,Omega1.c=Omega1.c)
if(y.type==3)
{caty=nlevels(y)}
if(y.type==4)
{zero=rep(0,nrow(y))#is.cen=ifelse(y[,2]==1,0,1) #censored or not
if(is.null(tmax))
tmax=max(y[,1], na.rm=T)+100
if(is.null(multi))
multi=TRUE} #maximum time to event
#Cen=ifelse(y[,2]==1,tmax,y[,1]) #censoring time
#y=ifelse(y[,2]==1,y[,1],NA)}
data0.1<- list (N=N,x=x,x1=x1,y=y,M1=M1,M2=M2,Omega=Omega,mu=mu,P=P,c1=c1,c2=c2,
muc=muc,Omegac=Omegac,nmc=nmc,mcov=mcov,mind=mind)
para=c("beta0","c","beta")
if(y.type==3)
{data0.1[['caty']]=caty}
if(y.type==4)
{data0.1[['zero']]=zero
#data0.1[['is.cen']]=is.cen
data0.1[['tmax']]=tmax
para=c(para,"lambda","r")}
# if(c1>1)
# {data0.1[['muc']]=rep(0,c1)
# data0.1[['Omegac']]=diag(0.000001,c1)
# } #data0.1[['c1']]=c1
if(p1>0)
{data0.1[['contm']]=contm
data0.1[['contm1']]=contm1
data0.1[['p1']]=p1
data0.1[['mu0.1']]=mu0.1
data0.1[['mu1.1']]=mu1.1
data0.1[['Omega0.1']]=Omega0.1
data0.1[['Omega1.1']]=Omega1.1
data0.1[['mu0.a']]=mu0.a
data0.1[['mu1.a']]=mu1.a
data0.1[['Omega0.a']]=Omega0.a
data0.1[['Omega1.a']]=Omega1.a
para=c(para,"alpha0","alpha1","alpha0.a","alpha1.a")
}
if(p2>0)
{data0.1[['binm']]=binm
data0.1[['p2']]=p2
data0.1[['mu0.b']]=mu0.b
data0.1[['mu1.b']]=mu1.b
data0.1[['Omega0.b']]=Omega0.b
data0.1[['Omega1.b']]=Omega1.b
para=c(para,"alpha0.b","alpha1.b")}
if(p3>0)
{data0.1[['cat1']]=cat1
data0.1[['cat2']]=cat2
data0.1[['catm']]=catm
data0.1[['p3']]=p3
data0.1[['mu0.c']]=mu0.c
data0.1[['mu1.c']]=mu1.c
data0.1[['Omega0.c']]=Omega0.c
data0.1[['Omega1.c']]=Omega1.c
para=c(para,"alpha0.c","alpha1.c")}
if(!is.null(cova)){
cv1=ncol(cova)
if(is.null(mucv))
mucv=rep(0,cv1)
if(is.null(Omegacv))
Omegacv=diag(preci,cv1)
data0.1[['cv1']]=cv1
data0.1[['mucv']]=mucv
data0.1[['Omegacv']]=Omegacv
data0.1[['cova']]=cova
para=c(para,"eta")
}
if(is.null(inits))
inits<- function(){list()}
if (y.type==1 & is.null(cova))
ytype="contnc"
else if(y.type==1 & !is.null(cova))
ytype="contc"
else if(y.type==2 & is.null(cova))
ytype="binnc"
else if(y.type==2 & !is.null(cova))
ytype="binc"
else if(y.type==3 & is.null(cova))
ytype="catnc"
else if(y.type==3 & !is.null(cova))
ytype="catc"
else if(y.type==4 & is.null(cova))
ytype="survnc"
else if(y.type==4 & !is.null(cova))
ytype="survc"
if (is.null(filename))
{filename=paste(tempdir(),"model.txt",sep="/")
writeLines(jags_model(contm=ifelse(p1==0,'ncontm', 'ycontm'),
binm=ifelse(p2==0, 'nbinm','ybinm'),
catm=ifelse(p3==0, 'ncatm','ycatm'),
cova=ifelse(is.null(cova),'ncova','ycova'),
ytype=ytype), filename)
}
med0<- jags(data0.1, inits,
model.file = filename, #"O:/My Documents/My Research/Research/Bayesian Mediation Analysis/codes/promis/bx_cy.txt",
parameters.to.save = para,
n.chains = n.chains, n.iter = n.iter, n.burnin=n.burnin, n.thin = n.thin)
#check the results
#calculate the mediation effects
N1=(n.iter-n.burnin)/n.thin #10000
#m.mcov=apply(mcov,2,mean,na.rm=T) #effects are calculated at the mean of mcov
#med0$BUGSoutput$sims.list$r=1/med0$BUGSoutput$sims.list$r
#attach(med0$BUGSoutput$sims.list)
if(y.type==3){
aie1=array(0,c(N1,p1+p2+p3,c1,caty-1))
ie1=array(0,dim=c(N1,N,p1+p2+p3,c1,caty-1))
tt2=1
de1=array(0,c(N1,c1,caty-1))
n.cont=1
aie2=array(0,c(N1,p1+p2+p3,c1,caty-1))
ie2=array(0,dim=c(N1,N,p1+p2+p3,c1,caty-1))
de2=array(0,c(N1,c1,caty-1))
te3=array(0,c(N1,N,caty-1))
ate3=array(0,c(N1,c1,caty-1))
omu3=ate3
ade3=ate3
aie3=array(0,c(N1,p1+p2+p3,c1,caty-1))
ie3=array(0,dim=c(N1,N,p1+p2+p3,c1,caty-1))
de3=array(0,c(N1,N,caty-1))
mu_M2=array(0,c(dim(M1),N1))
mu_M3=mu_M2
te4=array(0,c(N1,N,caty-1))
de4=te4
ade4=array(0,c(N1,c1,caty-1))
ie4=array(0,dim=c(N1,N,p1+p2+p3,c1,caty-1))
mu.M0=array(0,c(dim(M1),N1))
mu.M1=mu.M0
ate4=ade4
omu4=ate4
aie4=array(0,c(N1,p1+p2+p3,c1,caty-1))
for (l in 1:c1){
x1.temp=x1
x3.temp=x1
if(l%in%data0$contpred1.0){
x3.temp[,l]=x1[,l]+deltap[l]
}
else if(l%in%data0$binpred1.0)
{x1.temp[,l]=0
x3.temp[,l]=1
deltap[l]=1}
else{ #categorical predictor
for (i in 1:nrow(data0$catpred1.0))
if(l%in%(data0$catpred1.0[i,1]:data0$catpred1.0[i,2]))
{x1.temp[,data0$catpred1.0[i,1]:data0$catpred1.0[i,2]]=0
x3.temp[,data0$catpred1.0[i,1]:data0$catpred1.0[i,2]]=0
x3.temp[,l]=1}
deltap[l]=1
}
#method 1: the same for binary or continuous predictors
if(p1>0){
for(k in 1:(caty-1))
if (p1+p2+p3==1 & c1==1 & contm1[1,1]==contm1[1,2])
aie1[,contm[1],l,k]=med0$BUGSoutput$sims.list$alpha1[,1]*med0$BUGSoutput$sims.list$beta[,k,contm1[1,1]]#*mean(data0$m.cont.der[,contm3[1,1]]) since alpha1 is the coefficient of x on the change of f(M), beta is the coefficient of f(M) on the change of y
else
for (j in 1: p1)
aie1[,contm[j],l,k]=apply(as.matrix(med0$BUGSoutput$sims.list$alpha1[,contm1[j,1]:contm1[j,2],l])*as.matrix(med0$BUGSoutput$sims.list$beta[,k,contm1[j,1]:contm1[j,2]]),1,sum)#*mean(data0$m.cont.der[,contm3[j,1]])
}
if(p2>0){
if(p2==1 & c1==1) #since c1=1, the predictor can only be binary or continuous
{if (nmc==1)
{temp.x1=med0$BUGSoutput$sims.list$alpha0.b[,1]+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,1])%*%t(x1.temp)
temp.x3=med0$BUGSoutput$sims.list$alpha0.b[,1]+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,1])%*%t(x3.temp)}
else
{temp.x1=med0$BUGSoutput$sims.list$alpha0.b[,1,mind[binm[1],]]%*%t(mcov[,mind[binm[1],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,1])%*%t(x1.temp)
temp.x3=med0$BUGSoutput$sims.list$alpha0.b[,1,mind[binm[1],]]%*%t(mcov[,mind[binm[1],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,1])%*%t(x3.temp)}
for(j in 1:(caty-1)){
if(!1%in%data0$contpred1.0) #if the predictor is binary
ie1[,,binm[1],1,j]=med0$BUGSoutput$sims.list$beta[,j,binm1[1]]*(exp(temp.x3)/(1+exp(temp.x3))-exp(temp.x1)/(1+exp(temp.x1)))
else #if the predictor is continuous
ie1[,,binm[1],1,j]=med0$BUGSoutput$sims.list$alpha1.b[,1]*med0$BUGSoutput$sims.list$beta[,j,binm1[1]]*exp(temp.x1)/(1+exp(temp.x1))^2
aie1[,binm[1],l,j] <-apply(ie1[,,binm[1],1,j],1,mean)}
}
else
for (k in 1:p2){
if (nmc==1 & p2==1)
{temp.x1=med0$BUGSoutput$sims.list$alpha0.b[,k]+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,k,])%*%t(x1.temp)
temp.x3=med0$BUGSoutput$sims.list$alpha0.b[,k]+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,k,])%*%t(x3.temp)
}
else
{temp.x1=med0$BUGSoutput$sims.list$alpha0.b[,k,mind[binm[k],]]%*%t(mcov[,mind[binm[k],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,k,])%*%t(x1.temp)
temp.x3=med0$BUGSoutput$sims.list$alpha0.b[,k,mind[binm[k],]]%*%t(mcov[,mind[binm[k],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,k,])%*%t(x3.temp)
}
for(j in 1:(caty-1)){
if(!l%in%data0$contpred1.0) #if the predictor is binary or categorical
ie1[,,binm[k],l,j]=med0$BUGSoutput$sims.list$beta[,j,binm1[k]]*(exp(temp.x3)/(1+exp(temp.x3))-exp(temp.x1)/(1+exp(temp.x1)))
else if(data0$contpred1.0) #for continuous predictor
ie1[,,binm[1],l,j]=med0$BUGSoutput$sims.list$alpha1.b[,k,l]*med0$BUGSoutput$sims.list$beta[,j,binm1[k]]*exp(temp.x1)/(1+exp(temp.x1))^2
aie1[,binm[k],l,j] <- apply(ie1[,,binm[k],l,j],1,mean)}
}
}
if(p3>0){
for (j in 1:p3){
mu_Mc1<-array(0,c(N1,N,cat2[j]-1))
mu_Mc0<-array(0,c(N1,N,cat2[j]-1))
for (k in 2:cat2[j]){
mu_Mc0[,,k-1] <- exp(as.matrix(med0$BUGSoutput$sims.list$alpha0.c[,j,k-1,mind[catm[j],]])%*%t(mcov[,mind[catm[j],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.c[,j,k-1,])%*%t(x1.temp))
mu_Mc1[,,k-1] <- exp(as.matrix(med0$BUGSoutput$sims.list$alpha0.c[,j,k-1,mind[catm[j],]])%*%t(mcov[,mind[catm[j],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.c[,j,k-1,])%*%t(x3.temp))
}
sum_Mc1 <-apply(mu_Mc1,c(1,2),sum)+1
sum_Mc0 <-apply(mu_Mc0,c(1,2),sum)+1
if(l%in%data0$contpred1.0) #for continuous predictor
{tt.0=rep(0,N1)
for (k in 2:cat2[j])
tt.0=mu_Mc0[,,k-1]/sum_Mc0*med0$BUGSoutput$sims.list$alpha1.c[,j,k-1,l]
for (q1 in 1:(caty-1))
for (k in 2:cat2[j])
ie1[,,catm[j],l,q1]=ie1[,,catm[j],l,q1]+(mu_Mc0[,,k-1]/sum_Mc0)*med0$BUGSoutput$sims.list$beta[,q1,catm1[j,1]+k-2]*(med0$BUGSoutput$sims.list$alpha1.c[,j,k-1,l]-tt.0)
}
else #for binary or categorical predictor
for (q1 in 1:(caty-1))
for (k in 2:cat2[j])
ie1[,,catm[j],l,q1]=ie1[,,catm[j],l]+(mu_Mc1[,,k-1]/sum_Mc1-mu_Mc0[,,k-1]/sum_Mc0)*med0$BUGSoutput$sims.list$beta[,q1,catm1[j,1]+k-2]
for (q1 in 1:(caty-1))
aie1[,catm[j],l,q1]<-apply(ie1[,,catm[j],l,q1],1,mean)
}}
if(l%in%data0$contpred1)
{tt1=match(l,data0$contpred1)
for (q1 in 1:(caty-1))
de1[,l,q1]=as.matrix(med0$BUGSoutput$sims.list$c[,q1,data0$contpred3[tt1,1]:data0$contpred3[tt1,2]])%*%
apply(as.matrix(data0$pred.cont.der[,data0$contpred3[tt1,1]:data0$contpred3[tt1,2]]),2,mean)
tt2=tt2+data0$contpred3[tt1,2]-data0$contpred3[tt1,1]+1
}
else
{for (q1 in 1:(caty-1))
de1[,l,q1]=med0$BUGSoutput$sims.list$c[,q1,tt2]
tt2=tt2+1}
#method2: the same for binary and continuous predictors
if(p1>0){
if(p1==1 & c1==1 & contm1[1,1]==contm1[1,2])
{temp.M3=M1
temp.M3[,contm1[1,1]]=M3[,contm3[1,1]]
temp.mu1<-array(0,c(N1,N,caty))
temp.mu3<-temp.mu1
for(q1 in 2:caty)
{temp.mu1[,,q1]=med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x)+med0$BUGSoutput$sims.list$beta[,q1-1,]%*%t(M1)+matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,nrow(x))
temp.mu3[,,q1]=med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x)+med0$BUGSoutput$sims.list$beta[,q1-1,]%*%t(temp.M3)+matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,nrow(x))
if(!is.null(cova))
{temp.mu1[,,q1]=temp.mu1[,,q1]+med0$BUGSoutput$sims.list$eta[,q1-1,]%*%t(cova)
temp.mu3[,,q1]=temp.mu3[,,q1]+med0$BUGSoutput$sims.list$eta[,q1-1,]%*%t(cova)}}
temp.mu1=exp(temp.mu1)
temp.mu3=exp(temp.mu3)
temp.mu1.sum=apply(temp.mu1,c(1,2),sum)
temp.mu3.sum=apply(temp.mu3,c(1,2),sum)
for(q1 in 1:(caty-1))
{ie2[,,contm[1],l,q1]=(med0$BUGSoutput$sims.list$alpha1.a[,1]/deltam[contm[1]])*(temp.mu3[,,q1+1]/temp.mu3.sum-temp.mu1[,,q1+1]/temp.mu1.sum)
aie2[,contm[1],l,q1]=apply(ie2[,,contm[1],l,q1],1,mean,na.rm=T)}
}
else
for (j in 1:p1){
temp.M3=M1
temp.M3[,contm1[j,1]:contm1[j,2]]=M3[,contm3[j,1]:contm3[j,2]]
temp.mu1<-array(0,c(N1,N,caty))
temp.mu3<-temp.mu1
for(q1 in 2:caty)
{temp.mu1[,,q1]=med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x)+med0$BUGSoutput$sims.list$beta[,q1-1,]%*%t(M1)+matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,nrow(x))
temp.mu3[,,q1]=med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x)+med0$BUGSoutput$sims.list$beta[,q1-1,]%*%t(temp.M3)+matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,nrow(x))
if(!is.null(cova))
{temp.mu1[,,q1]=temp.mu1[,,q1]+med0$BUGSoutput$sims.list$eta[,q1-1,]%*%t(cova)
temp.mu3[,,q1]=temp.mu3[,,q1]+med0$BUGSoutput$sims.list$eta[,q1-1,]%*%t(cova)}}
temp.mu1=exp(temp.mu1)
temp.mu3=exp(temp.mu3)
temp.mu1.sum=apply(temp.mu1,c(1,2),sum)
temp.mu3.sum=apply(temp.mu3,c(1,2),sum)
for(q1 in 1:(caty-1))
{ie2[,,contm[j],l,q1]=(med0$BUGSoutput$sims.list$alpha1.a[,j,l]/deltam[contm[j]])*(temp.mu3[,,q1+1]/temp.mu3.sum-temp.mu1[,,q1+1]/temp.mu1.sum)
aie2[,contm[j],l,q1]=apply(ie2[,,contm[j],l,q1],1,mean,na.rm=T)}
}}
#for binary and categorical mediators, method 2 and method 1 are not the same
if(p2>0){
if(p2==1 & c1==1){
temp.M3=M1
temp.M1=M1
temp.M3[,binm1[1]]=1
temp.M1[,binm1[1]]=0
temp.mu1<-array(0,c(N1,N,caty))
temp.mu3<-temp.mu1
for(q1 in 2:caty)
{temp.mu1[,,q1]=med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x)+med0$BUGSoutput$sims.list$beta[,q1-1,]%*%t(temp.M1)+matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,nrow(x))
temp.mu3[,,q1]=med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x)+med0$BUGSoutput$sims.list$beta[,q1-1,]%*%t(temp.M3)+matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,nrow(x))
if(!is.null(cova))
{temp.mu1[,,q1]=temp.mu1[,,q1]+med0$BUGSoutput$sims.list$eta[,q1-1,]%*%t(cova)
temp.mu3[,,q1]=temp.mu3[,,q1]+med0$BUGSoutput$sims.list$eta[,q1-1,]%*%t(cova)}}
temp.mu1=exp(temp.mu1)
temp.mu3=exp(temp.mu3)
temp.mu1.sum=apply(temp.mu1,c(1,2),sum)
temp.mu3.sum=apply(temp.mu3,c(1,2),sum)
bpart=array(0,c(N1,N,caty))
for(q1 in 1:(caty-1))
bpart[,,q1]=temp.mu3[,,q1+1]/temp.mu3.sum-temp.mu1[,,q1+1]/temp.mu1.sum
if (nmc==1){
temp.x1=med0$BUGSoutput$sims.list$alpha0.b[,1]+matrix(med0$BUGSoutput$sims.list$alpha1.b[,1])%*%t(x1.temp)
temp.x3=med0$BUGSoutput$sims.list$alpha0.b[,1]+matrix(med0$BUGSoutput$sims.list$alpha1.b[,1])%*%t(x3.temp)
}
else
{temp.x1=med0$BUGSoutput$sims.list$alpha0.b[,1,mind[binm[1],]]%*%t(mcov[,mind[binm[1],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,1])%*%t(x1.temp)
temp.x3=med0$BUGSoutput$sims.list$alpha0.b[,1,mind[binm[1],]]%*%t(mcov[,mind[binm[1],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,1])%*%t(x3.temp)}
for(q1 in 1:(caty-1)){
ie2[,,binm[1],1,q1]=bpart[,,q1]*(exp(temp.x3)/(1+exp(temp.x3))-exp(temp.x1)/(1+exp(temp.x1)))/deltap[l]
aie2[,binm[1],1,q1] <-apply(ie2[,,binm[1],1,q1],1,mean,na.rm=T)}}
else
for (k in 1:p2){
temp.M3=M1
temp.M1=M1
temp.M3[,binm1[k]]=1
temp.M1[,binm1[k]]=0
temp.mu1<-array(0,c(N1,N,caty))
temp.mu3<-temp.mu1
for(q1 in 2:caty)
{temp.mu1[,,q1]=med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x)+med0$BUGSoutput$sims.list$beta[,q1-1,]%*%t(temp.M1)+matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,nrow(x))
temp.mu3[,,q1]=med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x)+med0$BUGSoutput$sims.list$beta[,q1-1,]%*%t(temp.M3)+matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,nrow(x))
if(!is.null(cova))
{temp.mu1[,,q1]=temp.mu1[,,q1]+med0$BUGSoutput$sims.list$eta[,q1-1,]%*%t(cova)
temp.mu3[,,q1]=temp.mu3[,,q1]+med0$BUGSoutput$sims.list$eta[,q1-1,]%*%t(cova)}}
temp.mu1=exp(temp.mu1)
temp.mu3=exp(temp.mu3)
temp.mu1.sum=apply(temp.mu1,c(1,2),sum)
temp.mu3.sum=apply(temp.mu3,c(1,2),sum)
bpart=array(0,c(N1,N,caty-1))
for(q1 in 1:(caty-1))
bpart[,,q1]=temp.mu3[,,q1+1]/temp.mu3.sum-temp.mu1[,,q1+1]/temp.mu1.sum
if(nmc==1 & p2==1){
temp.x1=med0$BUGSoutput$sims.list$alpha0.b[,k]+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,k,])%*%t(x1.temp)
temp.x3=med0$BUGSoutput$sims.list$alpha0.b[,k]+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,k,])%*%t(x3.temp)
}
else{
temp.x1=med0$BUGSoutput$sims.list$alpha0.b[,k,mind[binm[k],]]%*%t(mcov[,mind[binm[k],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,k,])%*%t(x1.temp)
temp.x3=med0$BUGSoutput$sims.list$alpha0.b[,k,mind[binm[k],]]%*%t(mcov[,mind[binm[k],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,k,])%*%t(x3.temp)
}
for(q1 in 1:(caty-1)){
ie2[,,binm[k],l,q1]=bpart[,,q1]*(exp(temp.x3)/(1+exp(temp.x3))-exp(temp.x1)/(1+exp(temp.x1)))/deltap[l]
aie2[,binm[k],l,q1] <- apply(ie2[,,binm[k],l,q1],1,mean,na.rm=T)}
}
}
if(p3>0){
for (j in 1:p3){
bpart=array(0,c(N1,N,cat2[j]-1,caty-1))
M1.temp=M1
M1.temp[,catm1[j,1]:catm1[j,2]]=0
for (k in catm1[j,1]:catm1[j,2]){
temp.M3=M1.temp
temp.M1=M1.temp
temp.M3[,k]=1
temp.mu1<-array(0,c(N1,N,caty))
temp.mu3<-temp.mu1
for(q1 in 2:caty)
{temp.mu1[,,q1]=med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x)+med0$BUGSoutput$sims.list$beta[,q1-1,]%*%t(temp.M1)+matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,nrow(x))
temp.mu3[,,q1]=med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x)+med0$BUGSoutput$sims.list$beta[,q1-1,]%*%t(temp.M3)+matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,nrow(x))
if(!is.null(cova))
{temp.mu1[,,q1]=temp.mu1[,,q1]+med0$BUGSoutput$sims.list$eta[,q1-1,]%*%t(cova)
temp.mu3[,,q1]=temp.mu3[,,q1]+med0$BUGSoutput$sims.list$eta[,q1-1,]%*%t(cova)}}
temp.mu1=exp(temp.mu1)
temp.mu3=exp(temp.mu3)
temp.mu1.sum=apply(temp.mu1,c(1,2),sum)
temp.mu3.sum=apply(temp.mu3,c(1,2),sum)
for(q1 in 1:(caty-1))
bpart[,,k-catm1[j,1]+1,q1]=exp(temp.mu3)/(1+exp(temp.mu3))-exp(temp.mu1)/(1+exp(temp.mu1))
}
mu_Mc1<-array(0,c(N1,N,cat2[j]-1))
mu_Mc0<-array(0,c(N1,N,cat2[j]-1))
for (k in 2:cat2[j]){
mu_Mc1[,,k-1] <- exp(as.matrix(med0$BUGSoutput$sims.list$alpha0.c[,j,k-1,mind[catm[j],]])%*%t(mcov[,mind[catm[j],]])+med0$BUGSoutput$sims.list$alpha1.c[,j,k-1,]%*%t(x3.temp))
mu_Mc0[,,k-1] <- exp(as.matrix(med0$BUGSoutput$sims.list$alpha0.c[,j,k-1,mind[catm[j],]])%*%t(mcov[,mind[catm[j],]])+med0$BUGSoutput$sims.list$alpha1.c[,j,k-1,]%*%t(x1.temp))
}
sum_Mc1 <-apply(mu_Mc1,c(1,2),sum)+1
sum_Mc0 <-apply(mu_Mc0,c(1,2),sum)+1
for(q1 in 1:(caty-1))
{for (k in 2:cat2[j])
ie2[,,catm[j],l,q1]=ie2[,,catm[j],l,q1]+(mu_Mc1[,,k-1]/sum_Mc1-mu_Mc0[,,k-1]/sum_Mc0)*bpart[,,k-1,q1]
aie2[,catm[j],l,q1]<-apply(ie2[,,catm[j],l,q1],1,mean,na.rm=T)}
}
}
temp.x1=x
temp.x3=x
if(l%in%data0$contpred1.0)
{tt1=match(l,data0$contpred1.0)
temp.x3[,data0$contpred2[tt1,1]:data0$contpred2[tt1,2]]=data0$pred3[,data0$contpred3[tt1,1]:data0$contpred3[tt1,2]]}
else if(l%in%data0$binpred1.0)
{tt1=match(l,data0$binpred1)
temp.x1[,data0$binpred2[tt1]]=0
temp.x3[,data0$binpred2[tt1]]=1
deltap[l]=1}
else
{for (i in 1:nrow(data0$catpred1.0))
if(l%in%(data0$catpred1.0[i,1]:data0$catpred1.0[i,2]))
tt1=i
temp.x1[,data0$catpred2[tt1,1]:data0$catpred2[tt1,2]]=0
temp.x3[,data0$catpred2[tt1,1]:data0$catpred2[tt1,2]]=0
temp.x3[,l]=1
deltap[l]=1}
temp.mu1<-array(0,c(N1,N,caty))
temp.mu3<-temp.mu1
for(q1 in 2:caty)
{temp.mu1[,,q1]=med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(temp.x1)+med0$BUGSoutput$sims.list$beta[,q1-1,]%*%t(M1)+matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,nrow(x))
temp.mu3[,,q1]=med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(temp.x3)+med0$BUGSoutput$sims.list$beta[,q1-1,]%*%t(M1)+matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,nrow(x))
if(!is.null(cova))
{temp.mu1[,,q1]=temp.mu1[,,q1]+med0$BUGSoutput$sims.list$eta[,q1-1,]%*%t(cova)
temp.mu3[,,q1]=temp.mu3[,,q1]+med0$BUGSoutput$sims.list$eta[,q1-1,]%*%t(cova)}}
temp.mu1=exp(temp.mu1)
temp.mu3=exp(temp.mu3)
temp.mu1.sum=apply(temp.mu1,c(1,2),sum)
temp.mu3.sum=apply(temp.mu3,c(1,2),sum)
for(q1 in 1:(caty-1))
{de2.1=(temp.mu3[,,q1+1]/temp.mu3.sum-temp.mu1[,,q1+1]/temp.mu1.sum)/deltap[l]
de2[,l,q1]=apply(de2.1,1,mean,na.rm=T)}
#method 3:parametric
#3.1. get M1(x) and M1(x+dx) for y
if(p1>0){
if(c1==1 & p1+p2+p3==1 & contm1[1,1]==contm1[1,2]){
if(nmc==1)
{mu_M2[,contm1[1,1],] <- med0$BUGSoutput$sims.list$alpha0[,contm1[1,1]]+as.matrix(med0$BUGSoutput$sims.list$alpha1[,contm1[1,1]])%*%t(x1.temp)
mu_M3[,contm1[1,1],] <- med0$BUGSoutput$sims.list$alpha0[,contm1[1,1]]+as.matrix(med0$BUGSoutput$sims.list$alpha1[,contm1[1,1]])%*%t(x3.temp)}
else
{mu_M2[,contm1[1,1],] <- med0$BUGSoutput$sims.list$alpha0[,contm1[1,1],mind[contm[1],]]%*%t(mcov[,mind[contm[1],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1[,contm1[1,1]])%*%t(x1.temp)
mu_M3[,contm1[1,1],] <- med0$BUGSoutput$sims.list$alpha0[,contm1[1,1],mind[contm[1],]]%*%t(mcov[,mind[contm[1],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1[,contm1[1,1]])%*%t(x3.temp)}
}
else
for (j in 1:p1){
for (k in contm1[j,1]:contm1[j,2]){
if(nmc==1 & p1+p2+p3==1)
{mu_M2[,k,] <- med0$BUGSoutput$sims.list$alpha0[,k]+med0$BUGSoutput$sims.list$alpha1[,k,l]%*%t(x1.temp)
mu_M3[,k,] <- med0$BUGSoutput$sims.list$alpha0[,k]+med0$BUGSoutput$sims.list$alpha1[,k,l]%*%t(x3.temp)}
else
{mu_M2[,k,] <- med0$BUGSoutput$sims.list$alpha0[,k,mind[contm[j],]]%*%t(mcov[,mind[contm[j],]])+med0$BUGSoutput$sims.list$alpha1[,k,]%*%t(x1.temp)
mu_M3[,k,] <- med0$BUGSoutput$sims.list$alpha0[,k,mind[contm[j],]]%*%t(mcov[,mind[contm[j],]])+med0$BUGSoutput$sims.list$alpha1[,k,]%*%t(x3.temp)}
}}}
if(p2>0){
if(p2==1 & c1==1)
{if(nmc==1)
{temp.x1=med0$BUGSoutput$sims.list$alpha0.b[,1]+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,1])%*%t(x1.temp)
temp.x3=med0$BUGSoutput$sims.list$alpha0.b[,1]+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,1])%*%t(x3.temp)}
else
{temp.x1=med0$BUGSoutput$sims.list$alpha0.b[,1,mind[binm[1],]]%*%t(mcov[,mind[binm[1],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,1])%*%t(x1.temp)
temp.x3=med0$BUGSoutput$sims.list$alpha0.b[,1,mind[binm[1],]]%*%t(mcov[,mind[binm[1],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,1])%*%t(x3.temp)}
mu_M2[,binm[1],] <- exp(temp.x1)/(1+exp(temp.x1))
mu_M3[,binm[1],] <- exp(temp.x3)/(1+exp(temp.x3))
}
else{
for (k in 1:p2){
if(nmc==1 & p2==1)
{temp.x1=med0$BUGSoutput$sims.list$alpha0.b[,k]+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,k,])%*%t(x1.temp)
temp.x3=med0$BUGSoutput$sims.list$alpha0.b[,k]+med0$BUGSoutput$sims.list$alpha1.b[,k,]%*%t(x3.temp)}
else
{temp.x1=med0$BUGSoutput$sims.list$alpha0.b[,k,mind[binm[k],]]%*%t(mcov[,mind[binm[k],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,k,])%*%t(x1.temp)
temp.x3=med0$BUGSoutput$sims.list$alpha0.b[,k,mind[binm[k],]]%*%t(mcov[,mind[binm[k],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,k,])%*%t(x3.temp)}
mu_M2[,binm1[k],] <- exp(temp.x1)/(1+exp(temp.x1))
mu_M3[,binm1[k],] <- exp(temp.x3)/(1+exp(temp.x3))}
}}
if(p3>0){
for (j in 1:p3){
mu_Mc1<-array(0,c(N1,N,cat2[j]-1))
mu_Mc0<-array(0,c(N1,N,cat2[j]-1))
for (k in 2:cat2[j]){
mu_Mc1[,,k-1] <- exp(as.matrix(med0$BUGSoutput$sims.list$alpha0.c[,j,k-1,mind[catm[j],]])%*%t(mcov[,mind[catm[j],]])+med0$BUGSoutput$sims.list$alpha1.c[,j,k-1,]%*%t(x3.temp))
mu_Mc0[,,k-1] <- exp(as.matrix(med0$BUGSoutput$sims.list$alpha0.c[,j,k-1,mind[catm[j],]])%*%t(mcov[,mind[catm[j],]])+med0$BUGSoutput$sims.list$alpha1.c[,j,k-1,]%*%t(x1.temp))
}
sum_Mc1 <-apply(mu_Mc1,c(1,2),sum)+1
sum_Mc0 <-apply(mu_Mc0,c(1,2),sum)+1
for (k in 2:cat2[j])
{mu_M2[,catm1[j,1]+k-2,]=mu_Mc0[,,k-1]/sum_Mc0
mu_M3[,catm1[j,1]+k-2,]=mu_Mc1[,,k-1]/sum_Mc1}
}}
#3.2. get x and dx for y
x1.temp1=x
x3.temp1=x
if(l%in%as.vector(data0$contpred1.0)){ #need the continuous predictor in its original format
i=match(l,data0$contpred1.0)
x3.temp1[,data0$contpred2[i,1]:data0$contpred2[i,2]]=data0$pred3[,data0$contpred3[i,1]:data0$contpred3[i,2]]
}
else if(l%in%data0$binpred1.0)
{i=match(l,data0$binpred1.0)
x1.temp1[,data0$binpred2[i]]=0
x3.temp1[,data0$binpred2[i]]=1
deltap[l]=1}
else{ #categorical predictor
for (i in 1:nrow(data0$catpred1.0))
if(l%in%(data0$catpred1.0[i,1]:data0$catpred1.0[i,2]))
{x1.temp1[,data0$catpred2[i,1]:data0$catpred2[i,2]]=0
x3.temp1[,data0$catpred2[i,1]:data0$catpred2[i,2]]=0
di=match(l,data0$catpred1.0[i,1]:data0$catpred1.0[i,2])
x3.temp1[,data0$catpred2[i,1]+di-1]=1}
deltap[l]=1
}
#3.3. get the total effect
mu_y0<-array(0,c(N1,N,caty))
mu_y1<-mu_y0
for(q1 in 2:caty)
{temp1=array(0,c(nrow(M1)+1,ncol(M1),N1))
temp1[2:(nrow(M1)+1),,]=mu_M2
temp1[1,,]=t(med0$BUGSoutput$sims.list$beta[,q1-1,])
temp2=temp1
temp2[2:(nrow(M1)+1),,]=mu_M3
if(is.null(cova))
{mu_y0[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x1.temp1) + t(apply(temp1,3,matrix.prod))
mu_y1[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x3.temp1) + t(apply(temp2,3,matrix.prod))}
else
{mu_y0[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x1.temp1) + med0$BUGSoutput$sims.list$eta[,q1-1,]%*%t(as.matrix(cova))+t(apply(temp1,3,matrix.prod))
mu_y1[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x3.temp1) + med0$BUGSoutput$sims.list$eta[,q1-1,]%*%t(as.matrix(cova))+t(apply(temp2,3,matrix.prod))
}} #get the linear part
mu_y0=exp(mu_y0)
mu_y1=exp(mu_y1)
mu_y0.sum=apply(mu_y0,c(1,2),sum)
mu_y1.sum=apply(mu_y1,c(1,2),sum)
for(q1 in 1:(caty-1)){
te3[,,q1]=(mu_y1[,,q1+1]/mu_y1.sum-mu_y0[,,q1+1]/mu_y0.sum)/deltap[l]
ate3[,l,q1]=apply(te3[,,q1],1,mean,na.rm=T)}
#3.4. calculate the ie
j1=sample(1:N,size=N*N1,replace=T)
j2=sample(1:N,size=N*N1,replace=T)
#3.4.1. continuous mediators
if(p1>0){
for (j in 1:p1){
mu_y0.2<-array(0,c(N1,N,caty))
mu_y1.2<-mu_y0.2
for(q1 in 2:caty)
{temp1.1=temp1
temp1.2=temp2
temp1.1[1,,]=t(med0$BUGSoutput$sims.list$beta[,q1-1,])
temp1.2[1,,]=t(med0$BUGSoutput$sims.list$beta[,q1-1,])
for (i in contm1[j,1]:contm1[j,2])
{temp1.1[2:(nrow(M1)+1),i,]=matrix(M1[j1,i],N,N1)
temp1.2[2:(nrow(M1)+1),i,]=matrix(M1[j1,i],N,N1)} #j2/j1
if(is.null(cova))
{mu_y0.2[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x1.temp1) +t(apply(temp1.1,3,matrix.prod))
mu_y1.2[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x3.temp1) +t(apply(temp1.2,3,matrix.prod)) }
else{
mu_y0.2[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x1.temp1) + med0$BUGSoutput$sims.list$eta[,q1-1,]%*%t(cova)+t(apply(temp1.1,3,matrix.prod))
mu_y1.2[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x3.temp1) + med0$BUGSoutput$sims.list$eta[,q1-1,]%*%t(cova)+t(apply(temp1.2,3,matrix.prod))}
}
mu_y0.2=exp(mu_y0.2)
mu_y1.2=exp(mu_y1.2)
mu_y0.2.sum=apply(mu_y0.2,c(1,2),sum)
mu_y1.2.sum=apply(mu_y1.2,c(1,2),sum)
for(q1 in 1:(caty-1))
ie3[,,contm[j],l,q1]<-te3[,,q1]-(mu_y1.2[,,q1+1]/mu_y1.2.sum-mu_y0.2[,,q1+1]/mu_y0.2.sum)/deltap[l]
}}
#3.4.2. binary mediators
if(p2>0){
for (k in 1:p2){
mu_y0.2<-array(0,c(N1,N,caty))
mu_y1.2<-mu_y0.2
for(q1 in 2:caty)
{temp1.1=temp1
temp1.2=temp2
temp1.1[1,,]=t(med0$BUGSoutput$sims.list$beta[,q1-1,])
temp1.2[1,,]=t(med0$BUGSoutput$sims.list$beta[,q1-1,])
temp1.1[2:(nrow(M1)+1),binm1[k],]=matrix(M1[j1,binm1[k]],N,N1)
temp1.2[2:(nrow(M1)+1),binm1[k],]=matrix(M1[j1,binm1[k]],N,N1) #j2/j1
if (is.null(cova)){
mu_y0.2[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x1.temp1) + t(apply(temp1.1,3,matrix.prod))
mu_y1.2[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x3.temp1) + t(apply(temp1.2,3,matrix.prod))}
else{
mu_y0.2[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x1.temp1) + med0$BUGSoutput$sims.list$eta[,q1-1,]%*%t(cova)+t(apply(temp1.1,3,matrix.prod))
mu_y1.2[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x3.temp1) + med0$BUGSoutput$sims.list$eta[,q1-1,]%*%t(cova)+t(apply(temp1.2,3,matrix.prod))}
}
mu_y0.2=exp(mu_y0.2)
mu_y1.2=exp(mu_y1.2)
mu_y0.2.sum=apply(mu_y0.2,c(1,2),sum)
mu_y1.2.sum=apply(mu_y1.2,c(1,2),sum)
for(q1 in 1:(caty-1))
ie3[,,binm[k],l,q1]<-te3[,,q1]-(mu_y1.2[,,q1+1]/mu_y1.2.sum-mu_y0.2[,,q1+1]/mu_y0.2.sum)/deltap[l]
}}
if(p3>0){
for (j in 1:p3){
mu_y0.2<-array(0,c(N1,N,caty))
mu_y1.2<-mu_y0.2
for(q1 in 2:caty)
{temp1.1=temp1
temp1.2=temp2
temp1.1[1,,]=t(med0$BUGSoutput$sims.list$beta[,q1-1,])
temp1.2[1,,]=t(med0$BUGSoutput$sims.list$beta[,q1-1,])
for (i in catm1[j,1]:catm1[j,2])
{temp1.1[2:(nrow(M1)+1),i,]=matrix(M1[j1,i],N,N1)
temp1.2[2:(nrow(M1)+1),i,]=matrix(M1[j1,i],N,N1)} #j2/j1
if(is.null(cova)){
mu_y0.2[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x1.temp1) + t(apply(temp1.1,3,matrix.prod))
mu_y1.2[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x3.temp1) + t(apply(temp1.2,3,matrix.prod))}
else{
mu_y0.2[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x1.temp1) + med0$BUGSoutput$sims.list$eta[,q1-1,]%*%t(cova)+t(apply(temp1.1,3,matrix.prod))
mu_y1.2[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x3.temp1) + med0$BUGSoutput$sims.list$eta[,q1-1,]%*%t(cova)+t(apply(temp1.2,3,matrix.prod))}
}
mu_y0.2=exp(mu_y0.2)
mu_y1.2=exp(mu_y1.2)
mu_y0.2.sum=apply(mu_y0.2,c(1,2),sum)
mu_y1.2.sum=apply(mu_y1.2,c(1,2),sum)
for(q1 in 1:(caty-1))
ie3[,,catm[j],l,q1]<-te3[,,q1]-(mu_y1.2[,,q1+1]/mu_y1.2.sum-mu_y0.2[,,q1+1]/mu_y0.2.sum)/deltap[l]
}}
aie3[,,l,]<-apply(array(ie3[,,,l,],c(N1,N,p1+p2+p3,caty-1)),c(1,3,4),mean,na.rm=T)
#3.5. Calculate the de
mu_y0.2<-array(0,c(N1,N,caty))
mu_y1.2<-mu_y0.2
for(q1 in 2:caty)
{temp1.1=temp1
temp1.2=temp2
temp1.1[1,,]=t(med0$BUGSoutput$sims.list$beta[,q1-1,])
temp1.2[1,,]=t(med0$BUGSoutput$sims.list$beta[,q1-1,])
for (i in 1:ncol(M1))
{temp1.1[2:(nrow(M1)+1),i,]=matrix(M1[j1,i],N,N1)
temp1.2[2:(nrow(M1)+1),i,]=matrix(M1[j1,i],N,N1)} #j2/j1
if(is.null(cova)){
mu_y0.2[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x1.temp1) + t(apply(temp1.1,3,matrix.prod))
mu_y1.2[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x3.temp1) + t(apply(temp1.2,3,matrix.prod))}
else{
mu_y0.2[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x1.temp1) + med0$BUGSoutput$sims.list$eta[,q1-1,]%*%t(cova)+t(apply(temp1.1,3,matrix.prod))
mu_y1.2[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x3.temp1) + med0$BUGSoutput$sims.list$eta[,q1-1,]%*%t(cova)+t(apply(temp1.2,3,matrix.prod))}
}
mu_y0.2=exp(mu_y0.2)
mu_y1.2=exp(mu_y1.2)
mu_y0.2.sum=apply(mu_y0.2,c(1,2),sum)
mu_y1.2.sum=apply(mu_y1.2,c(1,2),sum)
for(q1 in 1:(caty-1))
{de3[,,q1]<-(mu_y1.2[,,q1+1]/mu_y1.2.sum-mu_y0.2[,,q1+1]/mu_y0.2.sum)/deltap[l]
ade3[,l,q1]=apply(de3[,,q1],1,mean,na.rm=T)}
#method3: semi-parametric for binary or categorical predictors
if(!l%in%data0$contpred1.0){
if(!is.null(data0$binpred1.0))
{if (l%*%data0$binpred1.0)
{M.0=data.frame(M1[x1[,l]==0,])
y.0=y[x1[,l]==0]}
else
{for(i in 1:nrow(data0$catpred1.0))
if(l%in%(data0$catpred1.0[i,1]:data0$catpred1.0[i,2]))
tt1=i
M.0=data.frame(M1[apply(x1[,data0$catpred1.0[tt1,1]:data0$catpred1.0[[tt1,2]]]==1,1,sum)==0,])
y.0=y[apply(x1[,data0$catpred1.0[tt1,1]:data0$catpred1.0[[tt1,2]]]==1,1,sum)==0]}}
else
{for(i in 1:nrow(data0$catpred1.0))
if(l%in%(data0$catpred1.0[i,1]:data0$catpred1.0[i,2]))
tt1=i
M.0=data.frame(M1[apply(x1[,data0$catpred1.0[tt1,1]:data0$catpred1.0[[tt1,2]]]==1,1,sum)==0,])
y.0=y[apply(x1[,data0$catpred1.0[tt1,1]:data0$catpred1.0[[tt1,2]]]==1,1,sum)==0]}
y.1=y[x1[,l]==1]
M.1=data.frame(M1[x1[,l]==1,])
j1=sample(1:N,size=N*N1,replace=T)
j2=sample(1:N,size=N*N1,replace=T)
n3=nrow(M.0)
n4=nrow(M.1)
j3=sample(1:n3,size=N*N1,replace = T)
j4=sample(1:n4,size=N*N1,replace = T)
for (i in 1:ncol(M1))
{mu.M0[,i,]=matrix(M.0[j3,i],N,N1)
mu.M1[,i,]=matrix(M.1[j4,i],N,N1)}
#4.1. get the total effect
mu_y0=matrix(y.0[j3],N1,N)
mu_y1=matrix(y.1[j4],N1,N)
temp1=array(0,c(nrow(M1)+1,ncol(M1),N1))
temp1[2:(nrow(M1)+1),,]=mu.M0
temp2=temp1
temp2[2:(nrow(M1)+1),,]=mu.M1
temp.levels=levels(y)[-1]
for(q1 in 1:length(temp.levels))
{te4[,,q1]<- (mu_y1==temp.levels[q1])-(mu_y0==temp.levels[q1])
ate4[,l,q1]=apply(te4[,,q1],1,mean,na.rm=T)}
#4.2. Get the ies
#ie for continuous mediators
if(p1>0){
for (j in 1:p1){
mu_y0.2<-array(0,c(N1,N,caty))
mu_y1.2<-mu_y0.2
for(q1 in 2:caty)
{temp1.1=temp1
temp1.2=temp2
temp1.1[1,,]=t(med0$BUGSoutput$sims.list$beta[,q1-1,])
temp1.2[1,,]=t(med0$BUGSoutput$sims.list$beta[,q1-1,])
for (i in contm1[j,1]:contm1[j,2])
{temp1.1[2:(nrow(M1)+1),i,]=matrix(M1[j1,i],N,N1)
temp1.2[2:(nrow(M1)+1),i,]=matrix(M1[j1,i],N,N1)} #j2/j1
if(is.null(cova))
{mu_y0.2[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x1.temp1) +t(apply(temp1.1,3,matrix.prod))
mu_y1.2[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x3.temp1) +t(apply(temp1.2,3,matrix.prod)) }
else{
mu_y0.2[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x1.temp1) + med0$BUGSoutput$sims.list$eta[,q1-1,]%*%t(cova)+t(apply(temp1.1,3,matrix.prod))
mu_y1.2[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x3.temp1) + med0$BUGSoutput$sims.list$eta[,q1-1,]%*%t(cova)+t(apply(temp1.2,3,matrix.prod))}
}
mu_y0.2=exp(mu_y0.2)
mu_y1.2=exp(mu_y1.2)
mu_y0.2.sum=apply(mu_y0.2,c(1,2),sum)
mu_y1.2.sum=apply(mu_y1.2,c(1,2),sum)
for(q1 in 1:(caty-1))
ie4[,,contm[j],l,q1]<-te4[,,q1]-(mu_y1.2[,,q1+1]/mu_y1.2.sum-mu_y0.2[,,q1+1]/mu_y0.2.sum)/deltap[l]
}}
#ie for binary mediators
if(p2>0){
for (k in 1:p2){
mu_y0.2<-array(0,c(N1,N,caty))
mu_y1.2<-mu_y0.2
for(q1 in 2:caty)
{temp1.1=temp1
temp1.2=temp2
temp1.1[2:(nrow(M1)+1),binm1[k],]=matrix(M1[j1,binm1[k]],N,N1)
temp1.2[2:(nrow(M1)+1),binm1[k],]=matrix(M1[j1,binm1[k]],N,N1) #j2/j1
temp1.1[1,,]=t(med0$BUGSoutput$sims.list$beta[,q1-1,])
temp1.2[1,,]=t(med0$BUGSoutput$sims.list$beta[,q1-1,])
if (is.null(cova)){
mu_y0.2[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x1.temp1) + t(apply(temp1.1,3,matrix.prod))
mu_y1.2[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x3.temp1) + t(apply(temp1.2,3,matrix.prod))}
else{
mu_y0.2[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x1.temp1) + med0$BUGSoutput$sims.list$eta[,q1-1,]%*%t(cova)+t(apply(temp1.1,3,matrix.prod))
mu_y1.2[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x3.temp1) + med0$BUGSoutput$sims.list$eta[,q1-1,]%*%t(cova)+t(apply(temp1.2,3,matrix.prod))}
}
mu_y0.2=exp(mu_y0.2)
mu_y1.2=exp(mu_y1.2)
mu_y0.2.sum=apply(mu_y0.2,c(1,2),sum)
mu_y1.2.sum=apply(mu_y1.2,c(1,2),sum)
for(q1 in 1:(caty-1))
ie4[,,binm[k],l,q1]<-te4[,,q1]-(mu_y1.2[,,q1+1]/mu_y1.2.sum-mu_y0.2[,,q1+1]/mu_y0.2.sum)
}}
#ie for categorical mediators
if(p3>0){
for (j in 1:p3){
mu_y0.2<-array(0,c(N1,N,caty))
mu_y1.2<-mu_y0.2
for(q1 in 2:caty)
{temp1.1=temp1
temp1.2=temp2
temp1.1[1,,]=t(med0$BUGSoutput$sims.list$beta[,q1-1,])
temp1.2[1,,]=t(med0$BUGSoutput$sims.list$beta[,q1-1,])
for (i in catm1[j,1]:catm1[j,2])
{temp1.1[2:(nrow(M1)+1),i,]=matrix(M1[j1,i],N,N1)
temp1.2[2:(nrow(M1)+1),i,]=matrix(M1[j1,i],N,N1)} #j2/j1
if(is.null(cova)){
mu_y0.2[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x1.temp1) + t(apply(temp1.1,3,matrix.prod))
mu_y1.2[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x3.temp1) + t(apply(temp1.2,3,matrix.prod))}
else{
mu_y0.2[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x1.temp1) + med0$BUGSoutput$sims.list$eta[,q1-1,]%*%t(cova)+t(apply(temp1.1,3,matrix.prod))
mu_y1.2[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x3.temp1) + med0$BUGSoutput$sims.list$eta[,q1-1,]%*%t(cova)+t(apply(temp1.2,3,matrix.prod))}
}
mu_y0.2=exp(mu_y0.2)
mu_y1.2=exp(mu_y1.2)
mu_y0.2.sum=apply(mu_y0.2,c(1,2),sum)
mu_y1.2.sum=apply(mu_y1.2,c(1,2),sum)
for(q1 in 1:(caty-1))
ie4[,,catm[j],l,q1]<-te4[,,q1]-(mu_y1.2[,,q1+1]/mu_y1.2.sum-mu_y0.2[,,q1+1]/mu_y0.2.sum)
}}
aie4[,,l,]<-apply(array(ie4[,,,l,],c(N1,N,p1+p2+p3,caty-1)),c(1,3,4),mean,na.rm=T)
#4.3. Calculate the de
mu_y0.2<-array(0,c(N1,N,caty))
mu_y1.2<-mu_y0.2
for(q1 in 2:caty)
{temp1.1=temp1
temp1.2=temp2
for (i in 1:ncol(M1))
{temp1.1[2:(nrow(M1)+1),i,]=matrix(M1[j1,i],N,N1)
temp1.2[2:(nrow(M1)+1),i,]=matrix(M1[j1,i],N,N1)} #j2/j1
temp1.1[1,,]=t(med0$BUGSoutput$sims.list$beta[,q1-1,])
temp1.2[1,,]=t(med0$BUGSoutput$sims.list$beta[,q1-1,])
if(is.null(cova)){
mu_y0.2[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x1.temp1) + t(apply(temp1.1,3,matrix.prod))
mu_y1.2[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x3.temp1) + t(apply(temp1.2,3,matrix.prod))}
else{
mu_y0.2[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x1.temp1) + med0$BUGSoutput$sims.list$eta[,q1-1,]%*%t(cova)+t(apply(temp1.1,3,matrix.prod))
mu_y1.2[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x3.temp1) + med0$BUGSoutput$sims.list$eta[,q1-1,]%*%t(cova)+t(apply(temp1.2,3,matrix.prod))}
}
mu_y0.2=exp(mu_y0.2)
mu_y1.2=exp(mu_y1.2)
mu_y0.2.sum=apply(mu_y0.2,c(1,2),sum)
mu_y1.2.sum=apply(mu_y1.2,c(1,2),sum)
}
}
ate1=apply(aie1,c(1,3,4),sum)+de1
ate2=apply(aie2,c(1,3,4),sum)+de2
colnames(aie4)=colnames(M2)
colnames(aie1)=colnames(M2)
colnames(aie2)=colnames(M2)
colnames(aie3)=colnames(M2)
}
else{
aie1=array(0,c(N1,p1+p2+p3,c1))
ie1=array(0,dim=c(N1,N,p1+p2+p3,c1))
tt2=1
de1=NULL
n.cont=1
aie2=array(0,c(N1,p1+p2+p3,c1))
ie2=array(0,dim=c(N1,N,p1+p2+p3,c1))
de2=NULL
ate3=matrix(0,N1,c1)
omu3=ate3
ade3=matrix(0,N1,c1)
aie3=array(0,c(N1,p1+p2+p3,c1))
ie3=array(0,dim=c(N1,N,p1+p2+p3,c1))
mu_M2=array(0,c(dim(M1),N1))
mu_M3=mu_M2
ade4=matrix(0,N1,c1)
ie4=array(0,dim=c(N1,N,p1+p2+p3,c1))
mu.M0=array(0,c(dim(M1),N1))
mu.M1=mu.M0
ate4=matrix(0,N1,c1)
omu4=ate4
aie4=array(0,c(N1,p1+p2+p3,c1))
for (l in 1:c1){
x1.temp=x1
x3.temp=x1
if(l%in%data0$contpred1.0){
x3.temp[,l]=x1[,l]+deltap[l]
}
else if(l%in%data0$binpred1.0)
{x1.temp[,l]=0
x3.temp[,l]=1
deltap[l]=1}
else{ #categorical predictor
for (i in 1:nrow(data0$catpred1.0))
if(l%in%(data0$catpred1.0[i,1]:data0$catpred1.0[i,2]))
{x1.temp[,data0$catpred1.0[i,1]:data0$catpred1.0[i,2]]=0
x3.temp[,data0$catpred1.0[i,1]:data0$catpred1.0[i,2]]=0
x3.temp[,l]=1}
deltap[l]=1
}
#method 1: the same for binary or continuous predictors
if(p1>0){
if (p1+p2+p3==1 & c1==1 & contm1[1,1]==contm1[1,2])
aie1[,contm[1],l]=med0$BUGSoutput$sims.list$alpha1[,1]*med0$BUGSoutput$sims.list$beta[,contm1[1,1]]#*mean(data0$m.cont.der[,contm3[1,1]]) since alpha1 is the coefficient of x on the change of f(M), beta is the coefficient of f(M) on the change of y
else
for (j in 1: p1)
aie1[,contm[j],l]=apply(as.matrix(med0$BUGSoutput$sims.list$alpha1[,contm1[j,1]:contm1[j,2],l])*as.matrix(med0$BUGSoutput$sims.list$beta[,contm1[j,1]:contm1[j,2]]),1,sum)#*mean(data0$m.cont.der[,contm3[j,1]])
}
if(p2>0){
if(p2==1 & c1==1) #since c1=1, the predictor can only be binary or continuous
{if (nmc==1)
{temp.x1=med0$BUGSoutput$sims.list$alpha0.b[,1]+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,1])%*%t(x1.temp)
temp.x3=med0$BUGSoutput$sims.list$alpha0.b[,1]+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,1])%*%t(x3.temp)}
else
{temp.x1=med0$BUGSoutput$sims.list$alpha0.b[,1,mind[binm[1],]]%*%t(mcov[,mind[binm[1],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,1])%*%t(x1.temp)
temp.x3=med0$BUGSoutput$sims.list$alpha0.b[,1,mind[binm[1],]]%*%t(mcov[,mind[binm[1],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,1])%*%t(x3.temp)}
if(!1%in%data0$contpred1.0) #if the predictor is binary
ie1[,,binm[1],1]=med0$BUGSoutput$sims.list$beta[,binm1[1]]*(exp(temp.x3)/(1+exp(temp.x3))-exp(temp.x1)/(1+exp(temp.x1)))
else #if the predictor is continuous
ie1[,,binm[1],1]=med0$BUGSoutput$sims.list$alpha1.b[,1]*med0$BUGSoutput$sims.list$beta[,binm1[1]]*exp(temp.x1)/(1+exp(temp.x1))^2
aie1[,binm[1],l] <-apply(ie1[,,binm[1],1],1,mean)
}
else
for (k in 1:p2){
if (nmc==1 & p2==1)
{temp.x1=med0$BUGSoutput$sims.list$alpha0.b[,k]+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,k,])%*%t(x1.temp)
temp.x3=med0$BUGSoutput$sims.list$alpha0.b[,k]+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,k,])%*%t(x3.temp)
}
else
{temp.x1=med0$BUGSoutput$sims.list$alpha0.b[,k,mind[binm[k],]]%*%t(mcov[,mind[binm[k],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,k,])%*%t(x1.temp)
temp.x3=med0$BUGSoutput$sims.list$alpha0.b[,k,mind[binm[k],]]%*%t(mcov[,mind[binm[k],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,k,])%*%t(x3.temp)
}
if(!l%in%data0$contpred1.0) #if the predictor is binary or categorical
ie1[,,binm[k],l]=med0$BUGSoutput$sims.list$beta[,binm1[k]]*(exp(temp.x3)/(1+exp(temp.x3))-exp(temp.x1)/(1+exp(temp.x1)))
else if(data0$contpred1.0) #for continuous predictor
ie1[,,binm[1],l]=med0$BUGSoutput$sims.list$alpha1.b[,k,l]*med0$BUGSoutput$sims.list$beta[,binm1[k]]*exp(temp.x1)/(1+exp(temp.x1))^2
aie1[,binm[k],l] <- apply(ie1[,,binm[k],l],1,mean)
}
}
if(p3>0){
for (j in 1:p3){
mu_Mc1<-array(0,c(N1,N,cat2[j]-1))
mu_Mc0<-array(0,c(N1,N,cat2[j]-1))
for (k in 2:cat2[j]){
mu_Mc0[,,k-1] <- exp(as.matrix(med0$BUGSoutput$sims.list$alpha0.c[,j,k-1,mind[catm[j],]])%*%t(mcov[,mind[catm[j],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.c[,j,k-1,])%*%t(x1.temp))
mu_Mc1[,,k-1] <- exp(as.matrix(med0$BUGSoutput$sims.list$alpha0.c[,j,k-1,mind[catm[j],]])%*%t(mcov[,mind[catm[j],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.c[,j,k-1,])%*%t(x3.temp))
}
sum_Mc1 <-apply(mu_Mc1,c(1,2),sum)+1
sum_Mc0 <-apply(mu_Mc0,c(1,2),sum)+1
if(l%in%data0$contpred1.0) #for continuous predictor
{tt.0=rep(0,N1)
for (k in 2:cat2[j])
tt.0=mu_Mc0[,,k-1]/sum_Mc0*med0$BUGSoutput$sims.list$alpha1.c[,j,k-1,l]
for (k in 2:cat2[j])
ie1[,,catm[j],l]=ie1[,,catm[j],l]+(mu_Mc0[,,k-1]/sum_Mc0)*med0$BUGSoutput$sims.list$beta[,catm1[j,1]+k-2]*(med0$BUGSoutput$sims.list$alpha1.c[,j,k-1,l]-tt.0)
}
else #for binary or categorical predictor
for (k in 2:cat2[j])
ie1[,,catm[j],l]=ie1[,,catm[j],l]+(mu_Mc1[,,k-1]/sum_Mc1-mu_Mc0[,,k-1]/sum_Mc0)*med0$BUGSoutput$sims.list$beta[,catm1[j,1]+k-2]
aie1[,catm[j],l]<-apply(ie1[,,catm[j],l],1,mean)
}}
if(l%in%data0$contpred1)
{tt1=match(l,data0$contpred1)
de1=cbind(de1,as.matrix(med0$BUGSoutput$sims.list$c[,data0$contpred3[tt1,1]:data0$contpred3[tt1,2]])%*%
apply(as.matrix(data0$pred.cont.der[,data0$contpred3[tt1,1]:data0$contpred3[tt1,2]]),2,mean))
tt2=tt2+data0$contpred3[tt1,2]-data0$contpred3[tt1,1]+1
}
else
{de1=cbind(de1,med0$BUGSoutput$sims.list$c[,tt2])
tt2=tt2+1}
#method2: the same for binary and continuous predictors
if(p1>0){
if(y.type==1){
if(p1==1 & c1==1 & contm1[1,1]==contm1[1,2])
{ ie2[,,contm[1],l]=(med0$BUGSoutput$sims.list$alpha1.a[,1]/deltam[1])* #
(as.matrix(med0$BUGSoutput$sims.list$beta[,contm1[1,1]])%*%(M3[,contm3[1,1]]-M1[,contm1[1,1]]))
aie2[,contm[1],l]=apply(ie2[,,contm[1],l],1,mean,na.rm=T)
}
else
for (j in 1:p1){
ie2[,,contm[j],l]=(med0$BUGSoutput$sims.list$alpha1.a[,j,l]/deltam[j])* #a unit change in x, result in the unit change in m
(med0$BUGSoutput$sims.list$beta[,contm1[j,1]:contm1[j,2]]%*%t(M3[,contm3[j,1]:contm3[j,2]]-M1[,contm1[j,1]:contm1[j,2]]))
aie2[,contm[j],l]=apply(ie2[,,contm[j],l],1,mean,na.rm=T) #unit change in m, result in the unit change in y
}}
else if(y.type==2)
{if(p1==1 & c1==1 & contm1[1,1]==contm1[1,2])
{ temp.M3=M1
temp.M3[,contm1[1,1]]=M3[,contm3[1,1]]
temp.mu1=med0$BUGSoutput$sims.list$c%*%t(x)+med0$BUGSoutput$sims.list$beta%*%t(M1)+matrix(med0$BUGSoutput$sims.list$beta0,N1,nrow(x))
temp.mu3=med0$BUGSoutput$sims.list$c%*%t(x)+med0$BUGSoutput$sims.list$beta%*%t(temp.M3)+matrix(med0$BUGSoutput$sims.list$beta0,N1,nrow(x))
if(!is.null(cova))
{temp.mu1=temp.mu1+med0$BUGSoutput$sims.list$eta%*%t(cova)
temp.mu3=temp.mu3+med0$BUGSoutput$sims.list$eta%*%t(cova)}
ie2[,,contm[1],l]=(med0$BUGSoutput$sims.list$alpha1.a[,1]/deltam[1])*(exp(temp.mu3)/(1+exp(temp.mu3))-exp(temp.mu1)/(1+exp(temp.mu1)))
aie2[,contm[1],l]=apply(ie2[,,contm[1],l],1,mean,na.rm=T)
}
else
for (j in 1:p1){
temp.M3=M1
temp.M3[,contm1[j,1]:contm1[j,2]]=M3[,contm3[j,1]:contm3[j,2]]
temp.mu1=med0$BUGSoutput$sims.list$c%*%t(x)+med0$BUGSoutput$sims.list$beta%*%t(M1)+matrix(med0$BUGSoutput$sims.list$beta0,N1,nrow(x))
temp.mu3=med0$BUGSoutput$sims.list$c%*%t(x)+med0$BUGSoutput$sims.list$beta%*%t(temp.M3)+matrix(med0$BUGSoutput$sims.list$beta0,N1,nrow(x))
if(!is.null(cova))
{temp.mu1=temp.mu1+med0$BUGSoutput$sims.list$eta%*%t(cova)
temp.mu3=temp.mu3+med0$BUGSoutput$sims.list$eta%*%t(cova)}
ie2[,,contm[j],l]=(med0$BUGSoutput$sims.list$alpha1.a[,j,l]/deltam[j])*(exp(temp.mu3)/(1+exp(temp.mu3))-exp(temp.mu1)/(1+exp(temp.mu1)))
aie2[,contm[j],l]=apply(ie2[,,contm[j],l],1,mean,na.rm=T)
}}
else if(y.type==4)
{if(p1==1 & c1==1 & contm1[1,1]==contm1[1,2])
{temp.M3=M1
temp.M3[,contm1[1,1]]=M3[,contm3[1,1]]
temp.mu1=med0$BUGSoutput$sims.list$c%*%t(x)+med0$BUGSoutput$sims.list$beta%*%t(M1)
temp.mu3=med0$BUGSoutput$sims.list$c%*%t(x)+med0$BUGSoutput$sims.list$beta%*%t(temp.M3)
if(!is.null(cova))
{temp.mu1=temp.mu1+med0$BUGSoutput$sims.list$eta%*%t(cova)
temp.mu3=temp.mu3+med0$BUGSoutput$sims.list$eta%*%t(cova)}
vec=cbind(med0$BUGSoutput$sims.list$r,
as.vector(med0$BUGSoutput$sims.list$lambda^(-1/med0$BUGSoutput$sims.list$r))*(1/apply(exp(temp.mu3),2,expp, 1/med0$BUGSoutput$sims.list$r)))
tmean3=t(apply(vec,1,weib.trunc.mean,right=tmax))
#if(multi)
# tmean3=log(tmean3)
vec=cbind(med0$BUGSoutput$sims.list$r,
as.vector(med0$BUGSoutput$sims.list$lambda^(-1/med0$BUGSoutput$sims.list$r))*(1/apply(exp(temp.mu1),2,expp, 1/med0$BUGSoutput$sims.list$r)))
tmean1=t(apply(vec,1,weib.trunc.mean,right=tmax))
#if(multi)
# tmean1=log(tmean1)
#tmean3=ifelse(tmean3==Inf,tmax-runif(1,0,0.1),tmean3)
#tmean1=ifelse(tmean1==Inf,tmax-runif(1,0,0.1),tmean1)
ie2[,,contm[1],l]=as.vector(med0$BUGSoutput$sims.list$alpha1.a[,1]/deltam[1])*(tmean3-tmean1)
aie2[,contm[1],l]=apply(ie2[,,contm[1],l],1,mean,na.rm=T)
}
else
for (j in 1:p1){
temp.M3=M1
temp.M3[,contm1[j,1]:contm1[j,2]]=M3[,contm3[j,1]:contm3[j,2]]
temp.mu1=med0$BUGSoutput$sims.list$c%*%t(x)+med0$BUGSoutput$sims.list$beta%*%t(M1)
temp.mu3=med0$BUGSoutput$sims.list$c%*%t(x)+med0$BUGSoutput$sims.list$beta%*%t(temp.M3)
if(!is.null(cova))
{temp.mu1=temp.mu1+med0$BUGSoutput$sims.list$eta%*%t(cova)
temp.mu3=temp.mu3+med0$BUGSoutput$sims.list$eta%*%t(cova)}
vec=cbind(med0$BUGSoutput$sims.list$r,
as.vector(med0$BUGSoutput$sims.list$lambda^(-1/med0$BUGSoutput$sims.list$r))*(1/apply(exp(temp.mu3),2,expp, 1/med0$BUGSoutput$sims.list$r)))
tmean3=t(apply(vec,1,weib.trunc.mean,right=tmax))
#if(multi)
# tmean3=log(tmean3)
vec=cbind(med0$BUGSoutput$sims.list$r,
as.vector(med0$BUGSoutput$sims.list$lambda^(-1/med0$BUGSoutput$sims.list$r))*(1/apply(exp(temp.mu1),2,expp, 1/med0$BUGSoutput$sims.list$r)))
tmean1=t(apply(vec,1,weib.trunc.mean,right=tmax))
#if(multi)
# tmean1=log(tmean1)
#tmean3=ifelse(tmean3==Inf,tmax-runif(1,0,0.1),tmean3)
#tmean1=ifelse(tmean1==Inf,tmax-runif(1,0,0.1),tmean1)
ie2[,,contm[j],l]=as.vector(med0$BUGSoutput$sims.list$alpha1.a[,j,l]/deltam[j])*(tmean3-tmean1)
aie2[,contm[j],l]=apply(ie2[,,contm[j],l],1,mean)
}}
}
#for binary and categorical mediators, method 2 and method 1 are not the same
if(p2>0){
if(y.type==1){
if(p2==1 & c1==1){
if (nmc==1){
temp.x1=med0$BUGSoutput$sims.list$alpha0.b[,1]+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,1])%*%t(x1.temp)
temp.x3=med0$BUGSoutput$sims.list$alpha0.b[,1]+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,1])%*%t(x3.temp)
}
else
{temp.x1=med0$BUGSoutput$sims.list$alpha0.b[,1,mind[binm[1],]]%*%t(mcov[,mind[binm[1],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,1])%*%t(x1.temp)
temp.x3=med0$BUGSoutput$sims.list$alpha0.b[,1,mind[binm[1],]]%*%t(mcov[,mind[binm[1],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,1])%*%t(x3.temp)}
ie2[,,binm[1],1]=med0$BUGSoutput$sims.list$beta[,binm1[1]]*(exp(temp.x3)/(1+exp(temp.x3))-exp(temp.x1)/(1+exp(temp.x1)))/deltap[l]
aie2[,binm[1],1] <-apply(ie2[,,binm[1],1],1,mean,na.rm=T)}
else
for (k in 1:p2){
if(nmc==1 & p2==1){
temp.x1=med0$BUGSoutput$sims.list$alpha0.b[,k]+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,k,])%*%t(x1.temp)
temp.x3=med0$BUGSoutput$sims.list$alpha0.b[,k]+med0$BUGSoutput$sims.list$alpha1.b[,k,]%*%t(x3.temp)
}
else{
temp.x1=med0$BUGSoutput$sims.list$alpha0.b[,k,mind[binm[k],]]%*%t(mcov[,mind[binm[k],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,k,])%*%t(x1.temp)
temp.x3=med0$BUGSoutput$sims.list$alpha0.b[,k,mind[binm[k],]]%*%t(mcov[,mind[binm[k],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,k,])%*%t(x3.temp)
}
ie2[,,binm[k],l]=med0$BUGSoutput$sims.list$beta[,binm1[k]]*(exp(temp.x3)/(1+exp(temp.x3))-exp(temp.x1)/(1+exp(temp.x1)))/deltap[l]
aie2[,binm[k],l] <- apply(ie2[,,binm[k],l],1,mean,na.rm=T)
}}
else if(y.type==2){
if(p2==1 & c1==1){
temp.M3=M1
temp.M1=M1
temp.M3[,binm1[1]]=1
temp.M1[,binm1[1]]=0
temp.mu1=med0$BUGSoutput$sims.list$c%*%t(x)+med0$BUGSoutput$sims.list$beta%*%t(temp.M1)+matrix(med0$BUGSoutput$sims.list$beta0,N1,nrow(x))
temp.mu3=med0$BUGSoutput$sims.list$c%*%t(x)+med0$BUGSoutput$sims.list$beta%*%t(temp.M3)+matrix(med0$BUGSoutput$sims.list$beta0,N1,nrow(x))
if(!is.null(cova))
{temp.mu1=temp.mu1+med0$BUGSoutput$sims.list$eta%*%t(cova)
temp.mu3=temp.mu3+med0$BUGSoutput$sims.list$eta%*%t(cova)}
bpart=exp(temp.mu3)/(1+exp(temp.mu3))-exp(temp.mu1)/(1+exp(temp.mu1))
if (nmc==1){
temp.x1=med0$BUGSoutput$sims.list$alpha0.b[,1]+matrix(med0$BUGSoutput$sims.list$alpha1.b[,1])%*%t(x1.temp)
temp.x3=med0$BUGSoutput$sims.list$alpha0.b[,1]+matrix(med0$BUGSoutput$sims.list$alpha1.b[,1])%*%t(x3.temp)
}
else
{temp.x1=med0$BUGSoutput$sims.list$alpha0.b[,1,mind[binm[1],]]%*%t(mcov[,mind[binm[1],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,1])%*%t(x1.temp)
temp.x3=med0$BUGSoutput$sims.list$alpha0.b[,1,mind[binm[1],]]%*%t(mcov[,mind[binm[1],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,1])%*%t(x3.temp)}
ie2[,,binm[1],1]=bpart*(exp(temp.x3)/(1+exp(temp.x3))-exp(temp.x1)/(1+exp(temp.x1)))/deltap[l]
aie2[,binm[1],1] <-apply(ie2[,,binm[1],1],1,mean,na.rm=T)}
else
for (k in 1:p2){
temp.M3=M1
temp.M1=M1
temp.M3[,binm1[k]]=1
temp.M1[,binm1[k]]=0
temp.mu1=med0$BUGSoutput$sims.list$c%*%t(x)+med0$BUGSoutput$sims.list$beta%*%t(temp.M1)+matrix(med0$BUGSoutput$sims.list$beta0,N1,nrow(x))
temp.mu3=med0$BUGSoutput$sims.list$c%*%t(x)+med0$BUGSoutput$sims.list$beta%*%t(temp.M3)+matrix(med0$BUGSoutput$sims.list$beta0,N1,nrow(x))
if(!is.null(cova))
{temp.mu1=temp.mu1+med0$BUGSoutput$sims.list$eta%*%t(cova)
temp.mu3=temp.mu3+med0$BUGSoutput$sims.list$eta%*%t(cova)}
bpart=exp(temp.mu3)/(1+exp(temp.mu3))-exp(temp.mu1)/(1+exp(temp.mu1))
if(nmc==1 & p2==1){
temp.x1=med0$BUGSoutput$sims.list$alpha0.b[,k]+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,k,])%*%t(x1.temp)
temp.x3=med0$BUGSoutput$sims.list$alpha0.b[,k]+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,k,])%*%t(x3.temp)
}
else{
temp.x1=med0$BUGSoutput$sims.list$alpha0.b[,k,mind[binm[k],]]%*%t(mcov[,mind[binm[k],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,k,])%*%t(x1.temp)
temp.x3=med0$BUGSoutput$sims.list$alpha0.b[,k,mind[binm[k],]]%*%t(mcov[,mind[binm[k],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,k,])%*%t(x3.temp)
}
ie2[,,binm[k],l]=bpart*(exp(temp.x3)/(1+exp(temp.x3))-exp(temp.x1)/(1+exp(temp.x1)))/deltap[l]
aie2[,binm[k],l] <- apply(ie2[,,binm[k],l],1,mean,na.rm=T)
}}
else if(y.type==4){
if(p2==1 & c1==1){
temp.M3=M1
temp.M1=M1
temp.M3[,binm1[1]]=1
temp.M1[,binm1[1]]=0
temp.mu1=med0$BUGSoutput$sims.list$c%*%t(x)+med0$BUGSoutput$sims.list$beta%*%t(temp.M1)
temp.mu3=med0$BUGSoutput$sims.list$c%*%t(x)+med0$BUGSoutput$sims.list$beta%*%t(temp.M3)
if(!is.null(cova))
{temp.mu1=temp.mu1+med0$BUGSoutput$sims.list$eta%*%t(cova)
temp.mu3=temp.mu3+med0$BUGSoutput$sims.list$eta%*%t(cova)}
vec=cbind(med0$BUGSoutput$sims.list$r,
as.vector(med0$BUGSoutput$sims.list$lambda^(-1/med0$BUGSoutput$sims.list$r))*(1/apply(exp(temp.mu3),2,expp, 1/med0$BUGSoutput$sims.list$r)))
tmean3=t(apply(vec,1,weib.trunc.mean,right=tmax))
#if(multi)
# tmean3=log(tmean3)
vec=cbind(med0$BUGSoutput$sims.list$r,
as.vector(med0$BUGSoutput$sims.list$lambda^(-1/med0$BUGSoutput$sims.list$r))*(1/apply(exp(temp.mu1),2,expp, 1/med0$BUGSoutput$sims.list$r)))
tmean1=t(apply(vec,1,weib.trunc.mean,right=tmax))
#if(multi)
# tmean1=log(tmean1)
#tmean3=ifelse(tmean3==Inf,tmax-runif(1,0,0.1),tmean3)
#tmean1=ifelse(tmean1==Inf,tmax-runif(1,0,0.1),tmean1)
bpart=tmean3-tmean1
if (nmc==1){
temp.x1=med0$BUGSoutput$sims.list$alpha0.b[,1]+matrix(med0$BUGSoutput$sims.list$alpha1.b[,1])%*%t(x1.temp)
temp.x3=med0$BUGSoutput$sims.list$alpha0.b[,1]+matrix(med0$BUGSoutput$sims.list$alpha1.b[,1])%*%t(x3.temp)
}
else
{temp.x1=med0$BUGSoutput$sims.list$alpha0.b[,1,mind[binm[1],]]%*%t(mcov[,mind[binm[1],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,1])%*%t(x1.temp)
temp.x3=med0$BUGSoutput$sims.list$alpha0.b[,1,mind[binm[1],]]%*%t(mcov[,mind[binm[1],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,1])%*%t(x3.temp)}
ie2[,,binm[1],1]=bpart*(exp(temp.x3)/(1+exp(temp.x3))-exp(temp.x1)/(1+exp(temp.x1)))/deltap[l]
aie2[,binm[1],1] <-apply(ie2[,,binm[1],1],1,mean,na.rm=T)}
else
for (k in 1:p2){
temp.M3=M1
temp.M1=M1
temp.M3[,binm1[k]]=1
temp.M1[,binm1[k]]=0
temp.mu1=med0$BUGSoutput$sims.list$c%*%t(x)+med0$BUGSoutput$sims.list$beta%*%t(temp.M1)
temp.mu3=med0$BUGSoutput$sims.list$c%*%t(x)+med0$BUGSoutput$sims.list$beta%*%t(temp.M3)
if(!is.null(cova))
{temp.mu1=temp.mu1+med0$BUGSoutput$sims.list$eta%*%t(cova)
temp.mu3=temp.mu3+med0$BUGSoutput$sims.list$eta%*%t(cova)}
vec=cbind(med0$BUGSoutput$sims.list$r,
as.vector(med0$BUGSoutput$sims.list$lambda^(-1/med0$BUGSoutput$sims.list$r))*(1/apply(exp(temp.mu3),2,expp, 1/med0$BUGSoutput$sims.list$r)))
tmean3=t(apply(vec,1,weib.trunc.mean,right=tmax))
#if(multi)
# tmean3=log(tmean3)
vec=cbind(med0$BUGSoutput$sims.list$r,
as.vector(med0$BUGSoutput$sims.list$lambda^(-1/med0$BUGSoutput$sims.list$r))*(1/apply(exp(temp.mu1),2,expp, 1/med0$BUGSoutput$sims.list$r)))
tmean1=t(apply(vec,1,weib.trunc.mean,right=tmax))
#if(multi)
# tmean1=log(tmean1)
#tmean3=ifelse(tmean3==Inf,tmax-runif(1,0,0.1),tmean3)
#tmean1=ifelse(tmean1==Inf,tmax-runif(1,0,0.1),tmean1)
bpart=tmean3-tmean1
if(nmc==1 & p2==1){
temp.x1=med0$BUGSoutput$sims.list$alpha0.b[,k]+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,k,])%*%t(x1.temp)
temp.x3=med0$BUGSoutput$sims.list$alpha0.b[,k]+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,k,])%*%t(x3.temp)
}
else{
temp.x1=med0$BUGSoutput$sims.list$alpha0.b[,k,mind[binm[k],]]%*%t(mcov[,mind[binm[k],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,k,])%*%t(x1.temp)
temp.x3=med0$BUGSoutput$sims.list$alpha0.b[,k,mind[binm[k],]]%*%t(mcov[,mind[binm[k],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,k,])%*%t(x3.temp)
}
ie2[,,binm[k],l]=bpart*(exp(temp.x3)/(1+exp(temp.x3))-exp(temp.x1)/(1+exp(temp.x1)))/deltap[l]
aie2[,binm[k],l] <- apply(ie2[,,binm[k],l],1,mean,na.rm=T)
}}
}
if(p3>0){
if(y.type==1)
for (j in 1:p3){
mu_Mc1<-array(0,c(N1,N,cat2[j]-1))
mu_Mc0<-array(0,c(N1,N,cat2[j]-1))
for (k in 2:cat2[j]){
mu_Mc1[,,k-1] <- exp(as.matrix(med0$BUGSoutput$sims.list$alpha0.c[,j,k-1,mind[catm[j],]])%*%t(mcov[,mind[catm[j],]])+med0$BUGSoutput$sims.list$alpha1.c[,j,k-1,]%*%t(x3.temp))
mu_Mc0[,,k-1] <- exp(as.matrix(med0$BUGSoutput$sims.list$alpha0.c[,j,k-1,mind[catm[j],]])%*%t(mcov[,mind[catm[j],]])+med0$BUGSoutput$sims.list$alpha1.c[,j,k-1,]%*%t(x1.temp))
}
sum_Mc1 <-apply(mu_Mc1,c(1,2),sum)+1
sum_Mc0 <-apply(mu_Mc0,c(1,2),sum)+1
for (k in 2:cat2[j])
ie2[,,catm[j],l]=ie2[,,catm[j],l]+(mu_Mc1[,,k-1]/sum_Mc1-mu_Mc0[,,k-1]/sum_Mc0)*med0$BUGSoutput$sims.list$beta[,catm1[j,1]+k-2]/deltap[l]
aie2[,catm[j],l]<-apply(ie2[,,catm[j],l],1,mean,na.rm=T)
}
else if(y.type==2)
for (j in 1:p3){
bpart=array(0,c(N1,N,cat2[j]-1))
M1.temp=M1
M1.temp[,catm1[j,1]:catm1[j,2]]=0
for (k in catm1[j,1]:catm1[j,2]){
temp.M3=M1.temp
temp.M1=M1.temp
temp.M3[,k]=1
temp.mu1=med0$BUGSoutput$sims.list$c%*%t(x)+med0$BUGSoutput$sims.list$beta%*%t(temp.M1)+matrix(med0$BUGSoutput$sims.list$beta0,N1,nrow(x))
temp.mu3=med0$BUGSoutput$sims.list$c%*%t(x)+med0$BUGSoutput$sims.list$beta%*%t(temp.M3)+matrix(med0$BUGSoutput$sims.list$beta0,N1,nrow(x))
if(!is.null(cova))
{temp.mu1=temp.mu1+med0$BUGSoutput$sims.list$eta%*%t(cova)
temp.mu3=temp.mu3+med0$BUGSoutput$sims.list$eta%*%t(cova)}
bpart[,,k-catm1[j,1]+1]=exp(temp.mu3)/(1+exp(temp.mu3))-exp(temp.mu1)/(1+exp(temp.mu1))
}
mu_Mc1<-array(0,c(N1,N,cat2[j]-1))
mu_Mc0<-array(0,c(N1,N,cat2[j]-1))
for (k in 2:cat2[j]){
mu_Mc1[,,k-1] <- exp(as.matrix(med0$BUGSoutput$sims.list$alpha0.c[,j,k-1,mind[catm[j],]])%*%t(mcov[,mind[catm[j],]])+med0$BUGSoutput$sims.list$alpha1.c[,j,k-1,]%*%t(x3.temp))
mu_Mc0[,,k-1] <- exp(as.matrix(med0$BUGSoutput$sims.list$alpha0.c[,j,k-1,mind[catm[j],]])%*%t(mcov[,mind[catm[j],]])+med0$BUGSoutput$sims.list$alpha1.c[,j,k-1,]%*%t(x1.temp))
}
sum_Mc1 <-apply(mu_Mc1,c(1,2),sum)+1
sum_Mc0 <-apply(mu_Mc0,c(1,2),sum)+1
for (k in 2:cat2[j])
ie2[,,catm[j],l]=ie2[,,catm[j],l]+(mu_Mc1[,,k-1]/sum_Mc1-mu_Mc0[,,k-1]/sum_Mc0)*bpart[,,k-1]
# for (k in 1:N)
# ie2[,k,catm[j],l]=diag((diag(1/sum_Mc1[,k])%*%mu_Mc1[,k,]-diag(1/sum_Mc0[,k])%*%mu_Mc0[,k,])%*%t(bpart[,k,]))
aie2[,catm[j],l]<-apply(ie2[,,catm[j],l],1,mean,na.rm=T)
}
else if(y.type==4)
for (j in 1:p3){
bpart=array(0,c(N1,N,cat2[j]-1))
M1.temp=M1
M1.temp[,catm1[j,1]:catm1[j,2]]=0
for (k in catm1[j,1]:catm1[j,2]){
temp.M3=M1.temp
temp.M1=M1.temp
temp.M3[,k]=1
temp.mu1=med0$BUGSoutput$sims.list$c%*%t(x)+med0$BUGSoutput$sims.list$beta%*%t(temp.M1)
temp.mu3=med0$BUGSoutput$sims.list$c%*%t(x)+med0$BUGSoutput$sims.list$beta%*%t(temp.M3)
if(!is.null(cova))
{temp.mu1=temp.mu1+med0$BUGSoutput$sims.list$eta%*%t(cova)
temp.mu3=temp.mu3+med0$BUGSoutput$sims.list$eta%*%t(cova)}
vec=cbind(med0$BUGSoutput$sims.list$r,
as.vector(med0$BUGSoutput$sims.list$lambda^(-1/med0$BUGSoutput$sims.list$r))*(1/apply(exp(temp.mu3),2,expp, 1/med0$BUGSoutput$sims.list$r)))
tmean3=t(apply(vec,1,weib.trunc.mean,right=tmax))
#if(multi)
# tmean3=log(tmean3)
vec=cbind(med0$BUGSoutput$sims.list$r,
as.vector(med0$BUGSoutput$sims.list$lambda^(-1/med0$BUGSoutput$sims.list$r))*(1/apply(exp(temp.mu1),2,expp, 1/med0$BUGSoutput$sims.list$r)))
tmean1=t(apply(vec,1,weib.trunc.mean,right=tmax))
#if(multi)
# tmean1=log(tmean1)
#tmean3=ifelse(tmean3==Inf,tmax-runif(1,0,0.1),tmean3)
#tmean1=ifelse(tmean1==Inf,tmax-runif(1,0,0.1),tmean1)
bpart[,,k-catm1[j,1]+1]=tmean3-tmean1
# bpart[,,k-catm1[j,1]+1]=as.vector(gamma(1+1/med0$BUGSoutput$sims.list$r)/med0$BUGSoutput$sims.list$lambda^(1/med0$BUGSoutput$sims.list$r))*
# (1/apply(exp(temp.mu3),2,expp, 1/med0$BUGSoutput$sims.list$r)-1/apply(exp(temp.mu1),2,expp,1/med0$BUGSoutput$sims.list$r))
}
mu_Mc1<-array(0,c(N1,N,cat2[j]-1))
mu_Mc0<-array(0,c(N1,N,cat2[j]-1))
for (k in 2:cat2[j]){
mu_Mc1[,,k-1] <- exp(as.matrix(med0$BUGSoutput$sims.list$alpha0.c[,j,k-1,mind[catm[j],]])%*%t(mcov[,mind[catm[j],]])+med0$BUGSoutput$sims.list$alpha1.c[,j,k-1,]%*%t(x3.temp))
mu_Mc0[,,k-1] <- exp(as.matrix(med0$BUGSoutput$sims.list$alpha0.c[,j,k-1,mind[catm[j],]])%*%t(mcov[,mind[catm[j],]])+med0$BUGSoutput$sims.list$alpha1.c[,j,k-1,]%*%t(x1.temp))
}
sum_Mc1 <-apply(mu_Mc1,c(1,2),sum)+1
sum_Mc0 <-apply(mu_Mc0,c(1,2),sum)+1
for (k in 2:cat2[j])
ie2[,,catm[j],l]=ie2[,,catm[j],l]+(mu_Mc1[,,k-1]/sum_Mc1-mu_Mc0[,,k-1]/sum_Mc0)*bpart[,,k-1]
aie2[,catm[j],l]<-apply(ie2[,,catm[j],l],1,mean,na.rm=T)
}
}
if(y.type==2)
{temp.x1=x
temp.x3=x
if(l%in%data0$contpred1.0)
{tt1=match(l,data0$contpred1.0)
temp.x3[,data0$contpred2[tt1,1]:data0$contpred2[tt1,2]]=data0$pred3[,data0$contpred3[tt1,1]:data0$contpred3[tt1,2]]}
else if(l%in%data0$binpred1.0)
{tt1=match(l,data0$binpred1)
temp.x1[,data0$binpred2[tt1]]=0
temp.x3[,data0$binpred2[tt1]]=1
deltap[l]=1}
else
{for (i in 1:nrow(data0$catpred1.0))
if(l%in%(data0$catpred1.0[i,1]:data0$catpred1.0[i,2]))
tt1=i
temp.x1[,data0$catpred2[tt1,1]:data0$catpred2[tt1,2]]=0
temp.x3[,data0$catpred2[tt1,1]:data0$catpred2[tt1,2]]=0
temp.x3[,l]=1
deltap[l]=1}
temp.mu1=med0$BUGSoutput$sims.list$c%*%t(temp.x1)+med0$BUGSoutput$sims.list$beta%*%t(M1)+matrix(med0$BUGSoutput$sims.list$beta0,N1,nrow(x))
temp.mu3=med0$BUGSoutput$sims.list$c%*%t(temp.x3)+med0$BUGSoutput$sims.list$beta%*%t(M1)+matrix(med0$BUGSoutput$sims.list$beta0,N1,nrow(x))
if(!is.null(cova))
{temp.mu1=temp.mu1+med0$BUGSoutput$sims.list$eta%*%t(cova)
temp.mu3=temp.mu3+med0$BUGSoutput$sims.list$eta%*%t(cova)}
de2.1=(exp(temp.mu3)/(1+exp(temp.mu3))-exp(temp.mu1)/(1+exp(temp.mu1)))/deltap[l]
de2=cbind(de2,apply(de2.1,1,mean,na.rm=T))}
else if(y.type==4)
{temp.x1=x
temp.x3=x
if(l%in%data0$contpred1.0)
{tt1=match(l,data0$contpred1.0)
temp.x3[,data0$contpred2[tt1,1]:data0$contpred2[tt1,2]]=data0$pred3[,data0$contpred3[tt1,1]:data0$contpred3[tt1,2]]}
else if(l%in%data0$binpred1.0)
{tt1=match(l,data0$binpred1)
temp.x1[,data0$binpred2[tt1]]=0
temp.x3[,data0$binpred2[tt1]]=1
deltap[l]=1}
else
{for (i in 1:nrow(data0$catpred1.0))
if(l%in%(data0$catpred1.0[i,1]:data0$catpred1.0[i,2]))
tt1=i
temp.x1[,data0$catpred2[tt1,1]:data0$catpred2[tt1,2]]=0
temp.x3[,data0$catpred2[tt1,1]:data0$catpred2[tt1,2]]=0
temp.x3[,l]=1
deltap[l]=1}
temp.mu1=med0$BUGSoutput$sims.list$c%*%t(temp.x1)+med0$BUGSoutput$sims.list$beta%*%t(M1)
temp.mu3=med0$BUGSoutput$sims.list$c%*%t(temp.x3)+med0$BUGSoutput$sims.list$beta%*%t(M1)
if(!is.null(cova))
{temp.mu1=temp.mu1+med0$BUGSoutput$sims.list$eta%*%t(cova)
temp.mu3=temp.mu3+med0$BUGSoutput$sims.list$eta%*%t(cova)}
vec=cbind(med0$BUGSoutput$sims.list$r,
as.vector(med0$BUGSoutput$sims.list$lambda^(-1/med0$BUGSoutput$sims.list$r))*(1/apply(exp(temp.mu3),2,expp, 1/med0$BUGSoutput$sims.list$r)))
tmean3=t(apply(vec,1,weib.trunc.mean,right=tmax))
#if(multi)
# tmean3=log(tmean3)
vec=cbind(med0$BUGSoutput$sims.list$r,
as.vector(med0$BUGSoutput$sims.list$lambda^(-1/med0$BUGSoutput$sims.list$r))*(1/apply(exp(temp.mu1),2,expp, 1/med0$BUGSoutput$sims.list$r)))
tmean1=t(apply(vec,1,weib.trunc.mean,right=tmax))
#if(multi)
# tmean1=log(tmean1)
de2.1=(tmean3-tmean1)/deltap[l]
de2=cbind(de2,apply(de2.1,1,mean,na.rm=T))}
#method 3:parametric
#3.1. get M1(x) and M1(x+dx) for y
if(p1>0){
if(c1==1 & p1+p2+p3==1 & contm1[1,1]==contm1[1,2]){
if(nmc==1)
{mu_M2[,contm1[1,1],] <- med0$BUGSoutput$sims.list$alpha0[,contm1[1,1]]+as.matrix(med0$BUGSoutput$sims.list$alpha1[,contm1[1,1]])%*%t(x1.temp)
mu_M3[,contm1[1,1],] <- med0$BUGSoutput$sims.list$alpha0[,contm1[1,1]]+as.matrix(med0$BUGSoutput$sims.list$alpha1[,contm1[1,1]])%*%t(x3.temp)}
else
{mu_M2[,contm1[1,1],] <- med0$BUGSoutput$sims.list$alpha0[,contm1[1,1],mind[contm[1],]]%*%t(mcov[,mind[contm[1],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1[,contm1[1,1]])%*%t(x1.temp)
mu_M3[,contm1[1,1],] <- med0$BUGSoutput$sims.list$alpha0[,contm1[1,1],mind[contm[1],]]%*%t(mcov[,mind[contm[1],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1[,contm1[1,1]])%*%t(x3.temp)}
}
else
for (j in 1:p1){
for (k in contm1[j,1]:contm1[j,2]){
if(nmc==1 & p1+p2+p3==1)
{mu_M2[,k,] <- med0$BUGSoutput$sims.list$alpha0[,k]+med0$BUGSoutput$sims.list$alpha1[,k,l]%*%t(x1.temp)
mu_M3[,k,] <- med0$BUGSoutput$sims.list$alpha0[,k]+med0$BUGSoutput$sims.list$alpha1[,k,l]%*%t(x3.temp)}
else
{mu_M2[,k,] <- med0$BUGSoutput$sims.list$alpha0[,k,mind[contm[j],]]%*%t(mcov[,mind[contm[j],]])+med0$BUGSoutput$sims.list$alpha1[,k,]%*%t(x1.temp)
mu_M3[,k,] <- med0$BUGSoutput$sims.list$alpha0[,k,mind[contm[j],]]%*%t(mcov[,mind[contm[j],]])+med0$BUGSoutput$sims.list$alpha1[,k,]%*%t(x3.temp)}
}}}
if(p2>0){
if(p2==1 & c1==1)
{if(nmc==1)
{temp.x1=med0$BUGSoutput$sims.list$alpha0.b[,1]+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,1])%*%t(x1.temp)
temp.x3=med0$BUGSoutput$sims.list$alpha0.b[,1]+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,1])%*%t(x3.temp)}
else
{temp.x1=med0$BUGSoutput$sims.list$alpha0.b[,1,mind[binm[1],]]%*%t(mcov[,mind[binm[1],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,1])%*%t(x1.temp)
temp.x3=med0$BUGSoutput$sims.list$alpha0.b[,1,mind[binm[1],]]%*%t(mcov[,mind[binm[1],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,1])%*%t(x3.temp)}
mu_M2[,binm[1],] <- exp(temp.x1)/(1+exp(temp.x1))
mu_M3[,binm[1],] <- exp(temp.x3)/(1+exp(temp.x3))
}
else{
for (k in 1:p2){
if(nmc==1 & p2==1)
{temp.x1=med0$BUGSoutput$sims.list$alpha0.b[,k]+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,k,])%*%t(x1.temp)
temp.x3=med0$BUGSoutput$sims.list$alpha0.b[,k]+med0$BUGSoutput$sims.list$alpha1.b[,k,]%*%t(x3.temp)}
else
{temp.x1=med0$BUGSoutput$sims.list$alpha0.b[,k,mind[binm[k],]]%*%t(mcov[,mind[binm[k],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,k,])%*%t(x1.temp)
temp.x3=med0$BUGSoutput$sims.list$alpha0.b[,k,mind[binm[k],]]%*%t(mcov[,mind[binm[k],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,k,])%*%t(x3.temp)}
mu_M2[,binm1[k],] <- exp(temp.x1)/(1+exp(temp.x1))
mu_M3[,binm1[k],] <- exp(temp.x3)/(1+exp(temp.x3))}
}}
if(p3>0){
for (j in 1:p3){
mu_Mc1<-array(0,c(N1,N,cat2[j]-1))
mu_Mc0<-array(0,c(N1,N,cat2[j]-1))
for (k in 2:cat2[j]){
mu_Mc1[,,k-1] <- exp(as.matrix(med0$BUGSoutput$sims.list$alpha0.c[,j,k-1,mind[catm[j],]])%*%t(mcov[,mind[catm[j],]])+med0$BUGSoutput$sims.list$alpha1.c[,j,k-1,]%*%t(x3.temp))
mu_Mc0[,,k-1] <- exp(as.matrix(med0$BUGSoutput$sims.list$alpha0.c[,j,k-1,mind[catm[j],]])%*%t(mcov[,mind[catm[j],]])+med0$BUGSoutput$sims.list$alpha1.c[,j,k-1,]%*%t(x1.temp))
}
sum_Mc1 <-apply(mu_Mc1,c(1,2),sum)+1
sum_Mc0 <-apply(mu_Mc0,c(1,2),sum)+1
for (k in 2:cat2[j])
{mu_M2[,catm1[j,1]+k-2,]=mu_Mc0[,,k-1]/sum_Mc0
mu_M3[,catm1[j,1]+k-2,]=mu_Mc1[,,k-1]/sum_Mc1}
}}
#3.2. get x and dx for y
x1.temp1=x
x3.temp1=x
if(l%in%as.vector(data0$contpred1.0)){ #need the continuous predictor in its original format
i=match(l,data0$contpred1.0)
x3.temp1[,data0$contpred2[i,1]:data0$contpred2[i,2]]=data0$pred3[,data0$contpred3[i,1]:data0$contpred3[i,2]]
}
else if(l%in%data0$binpred1.0)
{i=match(l,data0$binpred1.0)
x1.temp1[,data0$binpred2[i]]=0
x3.temp1[,data0$binpred2[i]]=1
deltap[l]=1}
else{ #categorical predictor
for (i in 1:nrow(data0$catpred1.0))
if(l%in%(data0$catpred1.0[i,1]:data0$catpred1.0[i,2]))
{x1.temp1[,data0$catpred2[i,1]:data0$catpred2[i,2]]=0
x3.temp1[,data0$catpred2[i,1]:data0$catpred2[i,2]]=0
di=match(l,data0$catpred1.0[i,1]:data0$catpred1.0[i,2])
x3.temp1[,data0$catpred2[i,1]+di-1]=1}
deltap[l]=1
}
#3.3. get the total effect
temp1=array(0,c(nrow(M1)+1,ncol(M1),N1))
temp1[2:(nrow(M1)+1),,]=mu_M2
temp1[1,,]=t(med0$BUGSoutput$sims.list$beta)
temp2=temp1
temp2[2:(nrow(M1)+1),,]=mu_M3
if(is.null(cova))
{mu_y0<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x1.temp1) + t(apply(temp1,3,matrix.prod))
mu_y1<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x3.temp1) + t(apply(temp2,3,matrix.prod))
}
else
{mu_y0<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x1.temp1) + med0$BUGSoutput$sims.list$eta%*%t(as.matrix(cova))+t(apply(temp1,3,matrix.prod))
mu_y1<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x3.temp1) + med0$BUGSoutput$sims.list$eta%*%t(as.matrix(cova))+t(apply(temp2,3,matrix.prod))
} #get the linear part
if(y.type==2)
te3=(exp(mu_y1)/(1+exp(mu_y1))-exp(mu_y0)/(1+exp(mu_y0)))/deltap[l]
else if(y.type==1)
te3<- (mu_y1-mu_y0)/deltap[l]
else if (y.type==4)
{mu_y0<- mu_y0-matrix(med0$BUGSoutput$sims.list$beta0,N1,N)
mu_y1<- mu_y1-matrix(med0$BUGSoutput$sims.list$beta0,N1,N)
vec=cbind(med0$BUGSoutput$sims.list$r,
as.vector(med0$BUGSoutput$sims.list$lambda^(-1/med0$BUGSoutput$sims.list$r))*(1/apply(exp(mu_y1),2,expp, 1/med0$BUGSoutput$sims.list$r)))
tmean3=t(apply(vec,1,weib.trunc.mean,right=tmax))
vec=cbind(med0$BUGSoutput$sims.list$r,
as.vector(med0$BUGSoutput$sims.list$lambda^(-1/med0$BUGSoutput$sims.list$r))*(1/apply(exp(mu_y0),2,expp, 1/med0$BUGSoutput$sims.list$r)))
tmean1=t(apply(vec,1,weib.trunc.mean,right=tmax))
if(multi)
{tmean1=log(tmean1)
tmean3=log(tmean3)}
te3=(tmean3-tmean1)/deltap[l]
}
ate3[,l]=apply(te3,1,mean,na.rm=T)
if(y.type==4){
if(multi)
omu3[,l]=apply(exp(tmean1),1,mean,na.rm=T)
else
omu3[,l]=apply(tmean1,1,mean,na.rm=T)}
#3.4. calculate the ie
j1=sample(1:N,size=N*N1,replace=T)
j2=sample(1:N,size=N*N1,replace=T)
#3.4.1. continuous mediators
if(p1>0){
for (j in 1:p1){
temp1.1=temp1
temp1.2=temp2
for (i in contm1[j,1]:contm1[j,2])
{temp1.1[2:(nrow(M1)+1),i,]=matrix(M1[j1,i],N,N1)
temp1.2[2:(nrow(M1)+1),i,]=matrix(M1[j1,i],N,N1)} #j2/j1
if(is.null(cova))
{mu_y0.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x1.temp1) +t(apply(temp1.1,3,matrix.prod))
mu_y1.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x3.temp1) +t(apply(temp1.2,3,matrix.prod)) }
else{
mu_y0.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x1.temp1) + med0$BUGSoutput$sims.list$eta%*%t(cova)+t(apply(temp1.1,3,matrix.prod))
mu_y1.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x3.temp1) + med0$BUGSoutput$sims.list$eta%*%t(cova)+t(apply(temp1.2,3,matrix.prod))}
if(y.type==1)
ie3[,,contm[j],l]=te3-(mu_y1.2-mu_y0.2)/deltap[l]
else if(y.type==2)
ie3[,,contm[j],l]=te3-(exp(mu_y1.2)/(1+exp(mu_y1.2))-exp(mu_y0.2)/(1+exp(mu_y0.2)))/deltap[l]
else if (y.type==4)
{mu_y0.2<- mu_y0.2-matrix(med0$BUGSoutput$sims.list$beta0,N1,N)
mu_y1.2<- mu_y1.2-matrix(med0$BUGSoutput$sims.list$beta0,N1,N)
vec=cbind(med0$BUGSoutput$sims.list$r,
as.vector(med0$BUGSoutput$sims.list$lambda^(-1/med0$BUGSoutput$sims.list$r))*(1/apply(exp(mu_y1.2),2,expp, 1/med0$BUGSoutput$sims.list$r)))
tmean3=t(apply(vec,1,weib.trunc.mean,right=tmax))
if(multi)
tmean3=log(tmean3)
vec=cbind(med0$BUGSoutput$sims.list$r,
as.vector(med0$BUGSoutput$sims.list$lambda^(-1/med0$BUGSoutput$sims.list$r))*(1/apply(exp(mu_y0.2),2,expp, 1/med0$BUGSoutput$sims.list$r)))
tmean1=t(apply(vec,1,weib.trunc.mean,right=tmax))
if(multi)
tmean1=log(tmean1)
ie3[,,contm[j],l]<-te3-(tmean3-tmean1)/deltap[l]
}
}}
#3.4.2. binary mediators
if(p2>0){
for (k in 1:p2){
temp1.1=temp1
temp1.2=temp2
temp1.1[2:(nrow(M1)+1),binm1[k],]=matrix(M1[j1,binm1[k]],N,N1)
temp1.2[2:(nrow(M1)+1),binm1[k],]=matrix(M1[j1,binm1[k]],N,N1) #j2/j1
if (is.null(cova)){
mu_y0.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x1.temp1) + t(apply(temp1.1,3,matrix.prod))
mu_y1.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x3.temp1) + t(apply(temp1.2,3,matrix.prod))}
else{
mu_y0.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x1.temp1) + med0$BUGSoutput$sims.list$eta%*%t(cova)+t(apply(temp1.1,3,matrix.prod))
mu_y1.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x3.temp1) + med0$BUGSoutput$sims.list$eta%*%t(cova)+t(apply(temp1.2,3,matrix.prod))}
if(y.type==1)
ie3[,,binm[k],l]=te3-(mu_y1.2-mu_y0.2)/deltap[l]
else if(y.type==2)
ie3[,,binm[k],l]=te3-(exp(mu_y1.2)/(1+exp(mu_y1.2))-exp(mu_y0.2)/(1+exp(mu_y0.2)))/deltap[l]
else if (y.type==4)
{mu_y0.2<- mu_y0.2-matrix(med0$BUGSoutput$sims.list$beta0,N1,N)
mu_y1.2<- mu_y1.2-matrix(med0$BUGSoutput$sims.list$beta0,N1,N)
vec=cbind(med0$BUGSoutput$sims.list$r,
as.vector(med0$BUGSoutput$sims.list$lambda^(-1/med0$BUGSoutput$sims.list$r))*(1/apply(exp(mu_y1.2),2,expp, 1/med0$BUGSoutput$sims.list$r)))
tmean3=t(apply(vec,1,weib.trunc.mean,right=tmax))
if(multi)
tmean3=log(tmean3)
vec=cbind(med0$BUGSoutput$sims.list$r,
as.vector(med0$BUGSoutput$sims.list$lambda^(-1/med0$BUGSoutput$sims.list$r))*(1/apply(exp(mu_y0.2),2,expp, 1/med0$BUGSoutput$sims.list$r)))
tmean1=t(apply(vec,1,weib.trunc.mean,right=tmax))
if(multi)
tmean1=log(tmean1)
ie3[,,binm[k],l]<-te3-(tmean3-tmean1)/deltap[l]
}
}}
if(p3>0){
for (j in 1:p3){
temp1.1=temp1
temp1.2=temp2
for (i in catm1[j,1]:catm1[j,2])
{temp1.1[2:(nrow(M1)+1),i,]=matrix(M1[j1,i],N,N1)
temp1.2[2:(nrow(M1)+1),i,]=matrix(M1[j1,i],N,N1)} #j2/j1
if(is.null(cova)){
mu_y0.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x1.temp1) + t(apply(temp1.1,3,matrix.prod))
mu_y1.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x3.temp1) + t(apply(temp1.2,3,matrix.prod))}
else{
mu_y0.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x1.temp1) + med0$BUGSoutput$sims.list$eta%*%t(cova)+t(apply(temp1.1,3,matrix.prod))
mu_y1.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x3.temp1) + med0$BUGSoutput$sims.list$eta%*%t(cova)+t(apply(temp1.2,3,matrix.prod))}
if(y.type==1)
ie3[,,catm[j],l]=te3-(mu_y1.2-mu_y0.2)/deltap[l]
else if(y.type==2)
ie3[,,catm[j],l]=te3-(exp(mu_y1.2)/(1+exp(mu_y1.2))-exp(mu_y0.2)/(1+exp(mu_y0.2)))/deltap[l]
else if (y.type==4)
{mu_y0.2<- mu_y0.2-matrix(med0$BUGSoutput$sims.list$beta0,N1,N)
mu_y1.2<- mu_y1.2-matrix(med0$BUGSoutput$sims.list$beta0,N1,N)
vec=cbind(med0$BUGSoutput$sims.list$r,
as.vector(med0$BUGSoutput$sims.list$lambda^(-1/med0$BUGSoutput$sims.list$r))*(1/apply(exp(mu_y1.2),2,expp, 1/med0$BUGSoutput$sims.list$r)))
tmean3=t(apply(vec,1,weib.trunc.mean,right=tmax))
if(multi)
tmean3=log(tmean3)
vec=cbind(med0$BUGSoutput$sims.list$r,
as.vector(med0$BUGSoutput$sims.list$lambda^(-1/med0$BUGSoutput$sims.list$r))*(1/apply(exp(mu_y0.2),2,expp, 1/med0$BUGSoutput$sims.list$r)))
tmean1=t(apply(vec,1,weib.trunc.mean,right=tmax))
if(multi)
tmean1=log(tmean1)
ie3[,,catm[j],l]<-te3-(tmean3-tmean1)/deltap[l]
}
}}
aie3[,,l]<-apply(array(ie3[,,,l],c(N1,N,p1+p2+p3)),c(1,3),mean,na.rm=T)
#3.5. Calculate the de
temp1.1=temp1
temp1.2=temp2
for (i in 1:ncol(M1))
{temp1.1[2:(nrow(M1)+1),i,]=matrix(M1[j1,i],N,N1)
temp1.2[2:(nrow(M1)+1),i,]=matrix(M1[j1,i],N,N1)} #j2/j1
if(is.null(cova)){
mu_y0.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x1.temp1) + t(apply(temp1.1,3,matrix.prod))
mu_y1.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x3.temp1) + t(apply(temp1.2,3,matrix.prod))}
else{
mu_y0.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x1.temp1) + med0$BUGSoutput$sims.list$eta%*%t(cova)+t(apply(temp1.1,3,matrix.prod))
mu_y1.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x3.temp1) + med0$BUGSoutput$sims.list$eta%*%t(cova)+t(apply(temp1.2,3,matrix.prod))}
if(y.type==1)
de3=(mu_y1.2-mu_y0.2)/deltap[l]
else if(y.type==2)
de3=(exp(mu_y1.2)/(1+exp(mu_y1.2))-exp(mu_y0.2)/(1+exp(mu_y0.2)))/deltap[l]
else if (y.type==4)
{mu_y0.2<- mu_y0.2-matrix(med0$BUGSoutput$sims.list$beta0,N1,N)
mu_y1.2<- mu_y1.2-matrix(med0$BUGSoutput$sims.list$beta0,N1,N)
vec=cbind(med0$BUGSoutput$sims.list$r,
as.vector(med0$BUGSoutput$sims.list$lambda^(-1/med0$BUGSoutput$sims.list$r))*(1/apply(exp(mu_y1.2),2,expp, 1/med0$BUGSoutput$sims.list$r)))
tmean3=t(apply(vec,1,weib.trunc.mean,right=tmax))
if(multi)
tmean3=log(tmean3)
vec=cbind(med0$BUGSoutput$sims.list$r,
as.vector(med0$BUGSoutput$sims.list$lambda^(-1/med0$BUGSoutput$sims.list$r))*(1/apply(exp(mu_y0.2),2,expp, 1/med0$BUGSoutput$sims.list$r)))
tmean1=t(apply(vec,1,weib.trunc.mean,right=tmax))
if(multi)
tmean1=log(tmean1)
de3=(tmean3-tmean1)/deltap[l]
}
ade3[,l]=apply(de3,1,mean,na.rm=T)
#method3: semi-parametric for binary or categorical predictors
if(!l%in%data0$contpred1.0){
if(!is.null(data0$binpred1.0))
{if (l%*%data0$binpred1.0)
{M.0=data.frame(M1[x1[,l]==0,])
y.0=y[x1[,l]==0]}
else
{for(i in 1:nrow(data0$catpred1.0))
if(l%in%(data0$catpred1.0[i,1]:data0$catpred1.0[i,2]))
tt1=i
M.0=data.frame(M1[apply(x1[,data0$catpred1.0[tt1,1]:data0$catpred1.0[[tt1,2]]]==1,1,sum)==0,])
y.0=y[apply(x1[,data0$catpred1.0[tt1,1]:data0$catpred1.0[[tt1,2]]]==1,1,sum)==0]}}
else
{for(i in 1:nrow(data0$catpred1.0))
if(l%in%(data0$catpred1.0[i,1]:data0$catpred1.0[i,2]))
tt1=i
M.0=data.frame(M1[apply(x1[,data0$catpred1.0[tt1,1]:data0$catpred1.0[[tt1,2]]]==1,1,sum)==0,])
y.0=y[apply(x1[,data0$catpred1.0[tt1,1]:data0$catpred1.0[[tt1,2]]]==1,1,sum)==0]}
y.1=y[x1[,l]==1]
M.1=data.frame(M1[x1[,l]==1,])
j1=sample(1:N,size=N*N1,replace=T)
j2=sample(1:N,size=N*N1,replace=T)
n3=nrow(M.0)
n4=nrow(M.1)
j3=sample(1:n3,size=N*N1,replace = T)
j4=sample(1:n4,size=N*N1,replace = T)
for (i in 1:ncol(M1))
{mu.M0[,i,]=matrix(M.0[j3,i],N,N1)
mu.M1[,i,]=matrix(M.1[j4,i],N,N1)}
#4.1. get the total effect
mu_y0=matrix(y.0[j3],N1,N)
mu_y1=matrix(y.1[j4],N1,N)
temp1=array(0,c(nrow(M1)+1,ncol(M1),N1))
temp1[2:(nrow(M1)+1),,]=mu.M0
temp1[1,,]=t(med0$BUGSoutput$sims.list$beta)
temp2=temp1
temp2[2:(nrow(M1)+1),,]=mu.M1
if(is.null(cova))
{mu_y0.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x1.temp1) +t(apply(temp1,3,matrix.prod))
mu_y1.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x3.temp1) +t(apply(temp2,3,matrix.prod)) }
else{
mu_y0.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x1.temp1) + med0$BUGSoutput$sims.list$eta%*%t(cova)+t(apply(temp1,3,matrix.prod))
mu_y1.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x3.temp1) + med0$BUGSoutput$sims.list$eta%*%t(cova)+t(apply(temp2,3,matrix.prod))}
if(y.type==1)
te4<- (mu_y1.2-mu_y0.2)/deltap[l]
else if(y.type==2)
te4=(exp(mu_y1.2)/(1+exp(mu_y1.2))-exp(mu_y0.2)/(1+exp(mu_y0.2)))/deltap[l]
else if (y.type==4)
{mu_y0.2<- mu_y0.2-matrix(med0$BUGSoutput$sims.list$beta0,N1,N)
mu_y1.2<- mu_y1.2-matrix(med0$BUGSoutput$sims.list$beta0,N1,N)
vec=cbind(med0$BUGSoutput$sims.list$r,
as.vector(med0$BUGSoutput$sims.list$lambda^(-1/med0$BUGSoutput$sims.list$r))*(1/apply(exp(mu_y1.2),2,expp, 1/med0$BUGSoutput$sims.list$r)))
tmean3=t(apply(vec,1,weib.trunc.mean,right=tmax))
vec=cbind(med0$BUGSoutput$sims.list$r,
as.vector(med0$BUGSoutput$sims.list$lambda^(-1/med0$BUGSoutput$sims.list$r))*(1/apply(exp(mu_y0.2),2,expp, 1/med0$BUGSoutput$sims.list$r)))
tmean1=t(apply(vec,1,weib.trunc.mean,right=tmax))
if(multi)
{tmean1=log(tmean1)
tmean3=log(tmean3)}
te4=(tmean3-tmean1)/deltap[l]
}
ate4[,l]=apply(te4,1,mean,na.rm=T)
if(y.type==4){
if(multi)
omu4[,l]=apply(exp(tmean1),1,mean,na.rm=T)
else
omu4[,l]=apply(tmean1,1,mean,na.rm=T)}
#4.2. Get the ies
#ie for continuous mediators
if(p1>0){
for (j in 1:p1){
temp1.1=temp1
temp1.2=temp2
for (i in contm1[j,1]:contm1[j,2])
{temp1.1[2:(nrow(M1)+1),i,]=matrix(M1[j1,i],N,N1)
temp1.2[2:(nrow(M1)+1),i,]=matrix(M1[j1,i],N,N1)} #j2/j1
if(is.null(cova))
{mu_y0.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x1.temp1) +t(apply(temp1.1,3,matrix.prod))
mu_y1.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x3.temp1) +t(apply(temp1.2,3,matrix.prod)) }
else{
mu_y0.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x1.temp1) + med0$BUGSoutput$sims.list$eta%*%t(cova)+t(apply(temp1.1,3,matrix.prod))
mu_y1.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x3.temp1) + med0$BUGSoutput$sims.list$eta%*%t(cova)+t(apply(temp1.2,3,matrix.prod))}
if(y.type==1)
ie4[,,contm[j],l]=te4-(mu_y1.2-mu_y0.2)/deltap[l]
else if(y.type==2)
ie4[,,contm[j],l]=te4-(exp(mu_y1.2)/(1+exp(mu_y1.2))-exp(mu_y0.2)/(1+exp(mu_y0.2)))/deltap[l]
else if (y.type==4)
{mu_y0.2<- mu_y0.2-matrix(med0$BUGSoutput$sims.list$beta0,N1,N)
mu_y1.2<- mu_y1.2-matrix(med0$BUGSoutput$sims.list$beta0,N1,N)
vec=cbind(med0$BUGSoutput$sims.list$r,
as.vector(med0$BUGSoutput$sims.list$lambda^(-1/med0$BUGSoutput$sims.list$r))*(1/apply(exp(mu_y1.2),2,expp, 1/med0$BUGSoutput$sims.list$r)))
tmean3=t(apply(vec,1,weib.trunc.mean,right=tmax))
if(multi)
tmean3=log(tmean3)
vec=cbind(med0$BUGSoutput$sims.list$r,
as.vector(med0$BUGSoutput$sims.list$lambda^(-1/med0$BUGSoutput$sims.list$r))*(1/apply(exp(mu_y0.2),2,expp, 1/med0$BUGSoutput$sims.list$r)))
tmean1=t(apply(vec,1,weib.trunc.mean,right=tmax))
if(multi)
tmean1=log(tmean1)
ie4[,,contm[j],l]=te4-(tmean3-tmean1)/deltap[l]
}
}}
#ie for binary mediators
if(p2>0){
for (k in 1:p2){
temp1.1=temp1
temp1.2=temp2
temp1.1[2:(nrow(M1)+1),binm1[k],]=matrix(M1[j1,binm1[k]],N,N1)
temp1.2[2:(nrow(M1)+1),binm1[k],]=matrix(M1[j1,binm1[k]],N,N1) #j2/j1
if (is.null(cova)){
mu_y0.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x1.temp1) + t(apply(temp1.1,3,matrix.prod))
mu_y1.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x3.temp1) + t(apply(temp1.2,3,matrix.prod))}
else{
mu_y0.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x1.temp1) + med0$BUGSoutput$sims.list$eta%*%t(cova)+t(apply(temp1.1,3,matrix.prod))
mu_y1.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x3.temp1) + med0$BUGSoutput$sims.list$eta%*%t(cova)+t(apply(temp1.2,3,matrix.prod))}
if(y.type==1)
ie4[,,binm[k],l]=te4-(mu_y1.2-mu_y0.2)/deltap[l]
else if(y.type==2)
ie4[,,binm[k],l]=te4-(exp(mu_y1.2)/(1+exp(mu_y1.2))-exp(mu_y0.2)/(1+exp(mu_y0.2)))/deltap[l]
else if (y.type==4)
{mu_y0.2<- mu_y0.2-matrix(med0$BUGSoutput$sims.list$beta0,N1,N)
mu_y1.2<- mu_y1.2-matrix(med0$BUGSoutput$sims.list$beta0,N1,N)
vec=cbind(med0$BUGSoutput$sims.list$r,
as.vector(med0$BUGSoutput$sims.list$lambda^(-1/med0$BUGSoutput$sims.list$r))*(1/apply(exp(mu_y1.2),2,expp, 1/med0$BUGSoutput$sims.list$r)))
tmean3=t(apply(vec,1,weib.trunc.mean,right=tmax))
if(multi)
tmean3=log(tmean3)
vec=cbind(med0$BUGSoutput$sims.list$r,
as.vector(med0$BUGSoutput$sims.list$lambda^(-1/med0$BUGSoutput$sims.list$r))*(1/apply(exp(mu_y0.2),2,expp, 1/med0$BUGSoutput$sims.list$r)))
tmean1=t(apply(vec,1,weib.trunc.mean,right=tmax))
if(multi)
tmean1=log(tmean1)
ie4[,,binm[k],l]=te4-(tmean3-tmean1)/deltap[l]
}
}}
#ie for categorical mediators
if(p3>0){
for (j in 1:p3){
temp1.1=temp1
temp1.2=temp2
for (i in catm1[j,1]:catm1[j,2])
{temp1.1[2:(nrow(M1)+1),i,]=matrix(M1[j1,i],N,N1)
temp1.2[2:(nrow(M1)+1),i,]=matrix(M1[j1,i],N,N1)} #j2/j1
if(is.null(cova)){
mu_y0.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x1.temp1) + t(apply(temp1.1,3,matrix.prod))
mu_y1.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x3.temp1) + t(apply(temp1.2,3,matrix.prod))}
else{
mu_y0.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x1.temp1) + med0$BUGSoutput$sims.list$eta%*%t(cova)+t(apply(temp1.1,3,matrix.prod))
mu_y1.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x3.temp1) + med0$BUGSoutput$sims.list$eta%*%t(cova)+t(apply(temp1.2,3,matrix.prod))}
if(y.type==1)
ie4[,,catm[j],l]=te4-(mu_y1.2-mu_y0.2)/deltap[l]
else if(y.type==2)
ie4[,,catm[j],l]=te4-(exp(mu_y1.2)/(1+exp(mu_y1.2))-exp(mu_y0.2)/(1+exp(mu_y0.2)))/deltap[l]
else if (y.type==4)
{mu_y0.2<- mu_y0.2-matrix(med0$BUGSoutput$sims.list$beta0,N1,N)
mu_y1.2<- mu_y1.2-matrix(med0$BUGSoutput$sims.list$beta0,N1,N)
vec=cbind(med0$BUGSoutput$sims.list$r,
as.vector(med0$BUGSoutput$sims.list$lambda^(-1/med0$BUGSoutput$sims.list$r))*(1/apply(exp(mu_y1.2),2,expp, 1/med0$BUGSoutput$sims.list$r)))
tmean3=t(apply(vec,1,weib.trunc.mean,right=tmax))
if(multi)
tmean3=log(tmean3)
vec=cbind(med0$BUGSoutput$sims.list$r,
as.vector(med0$BUGSoutput$sims.list$lambda^(-1/med0$BUGSoutput$sims.list$r))*(1/apply(exp(mu_y0.2),2,expp, 1/med0$BUGSoutput$sims.list$r)))
tmean1=t(apply(vec,1,weib.trunc.mean,right=tmax))
if(multi)
tmean1=log(tmean1)
ie4[,,catm[j],l]=te4-(tmean3-tmean1)/deltap[l]
}
}}
aie4[,,l]<-apply(array(ie4[,,,l],c(N1,N,p1+p2+p3)),c(1,3),mean,na.rm=T)
#4.3. Calculate the de
temp1.1=temp1
temp1.2=temp2
for (i in 1:ncol(M1))
{temp1.1[2:(nrow(M1)+1),i,]=matrix(M1[j1,i],N,N1)
temp1.2[2:(nrow(M1)+1),i,]=matrix(M1[j1,i],N,N1)} #j2/j1
if(is.null(cova)){
mu_y0.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x1.temp1) + t(apply(temp1.1,3,matrix.prod))
mu_y1.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x3.temp1) + t(apply(temp1.2,3,matrix.prod))}
else{
mu_y0.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x1.temp1) + med0$BUGSoutput$sims.list$eta%*%t(cova)+t(apply(temp1.1,3,matrix.prod))
mu_y1.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x3.temp1) + med0$BUGSoutput$sims.list$eta%*%t(cova)+t(apply(temp1.2,3,matrix.prod))}
if(y.type==1)
de4=(mu_y1.2-mu_y0.2)/deltap[l]
else if(y.type==2)
de4=(exp(mu_y1.2)/(1+exp(mu_y1.2))-exp(mu_y0.2)/(1+exp(mu_y0.2)))/deltap[l]
else if (y.type==4)
{mu_y0.2<- mu_y0.2-matrix(med0$BUGSoutput$sims.list$beta0,N1,N)
mu_y1.2<- mu_y1.2-matrix(med0$BUGSoutput$sims.list$beta0,N1,N)
vec=cbind(med0$BUGSoutput$sims.list$r,
as.vector(med0$BUGSoutput$sims.list$lambda^(-1/med0$BUGSoutput$sims.list$r))*(1/apply(exp(mu_y1.2),2,expp, 1/med0$BUGSoutput$sims.list$r)))
tmean3=t(apply(vec,1,weib.trunc.mean,right=tmax))
if(multi)
tmean3=log(tmean3)
vec=cbind(med0$BUGSoutput$sims.list$r,
as.vector(med0$BUGSoutput$sims.list$lambda^(-1/med0$BUGSoutput$sims.list$r))*(1/apply(exp(mu_y0.2),2,expp, 1/med0$BUGSoutput$sims.list$r)))
tmean1=t(apply(vec,1,weib.trunc.mean,right=tmax))
if(multi)
tmean1=log(tmean1)
de4=(tmean3-tmean1)/deltap[l]
}
ade4[,l]=apply(de4,1,mean,na.rm=T)
}
}
if(y.type==1)
de2=de1
ate1=apply(aie1,c(1,3),sum)+de1
ate2=apply(aie2,c(1,3),sum)+de2
colnames(aie4)=colnames(M2)
colnames(aie1)=colnames(M2)
colnames(aie2)=colnames(M2)
colnames(aie3)=colnames(M2)
}
#detach(med0$BUGSoutput$sims.list)
result=list(aie1=aie1, ade1=de1, ate1=ate1, aie2=aie2, ade2=de2, ate2=ate2,
aie3=aie3, ade3=de3, ate3=ate3, aie4=aie4, ade4=ade4, ate4=ate4,
sims.list=med0,data0=data0,omu3=omu3,omu4=omu4) #ie2=ie2, med0$BUGSoutput$sims.list
class(result)='bma.bx.cy'
return(result)
}
summary.bma.bx.cy<-function(object, ..., plot= TRUE, RE=TRUE, quant=c(0.025, 0.25, 0.5, 0.75,0.975),digit=4,method=3)
{
summary.med<-function(vec,qua=quant, digit=digit)
{c(mean=mean(vec,na.rm=T),sd=sd(vec,na.rm=T),quantile(vec,qua,na.rm=T))
}
summary.med.re<-function(vec,vec1,qua=quant, digit=digit)
{vec=vec/vec1
c(mean=mean(vec,na.rm=T),sd=sd(vec,na.rm=T),quantile(vec,qua,na.rm=T))}
if(object$data0$y_type!=3){
result1<-array(0,c(7,2+ncol(object$aie1),ncol(object$ade1)))
result1.re<-array(0,c(7,1+ncol(object$aie1),ncol(object$ade1)))
result2<-result1
result2.re<-result1.re
result3<-result1
result3.re<-result1.re
result4<-result1
result4.re<-result1.re
for(j in 1:ncol(object$ade1)){
result1[,,j]<-apply(cbind(TE=object$ate1[,j],DE=object$ade1[,j],object$aie1[,,j]),2,summary.med)
result1.re[,,j]<-apply(cbind(DE=object$ade1[,j],object$aie1[,,j]),2,summary.med.re,object$ate1[,j])
result2[,,j]<-apply(cbind(TE=object$ate2[,j],DE=object$ade2[,j],object$aie2[,,j]),2,summary.med)
result2.re[,,j]<-apply(cbind(DE=object$ade2[,j],object$aie2[,,j]),2,summary.med.re,object$ate2[,j])
result3[,,j]<-apply(cbind(TE=object$ate3[,j],DE=object$ade3[,j],object$aie3[,,j]),2,summary.med)
result3.re[,,j]<-apply(cbind(DE=object$ade3[,j],object$aie3[,,j]),2,summary.med.re,object$ate3[,j])
result4[,,j]<-apply(cbind(TE=object$ate4[,j],DE=object$ade4[,j],object$aie4[,,j]),2,summary.med)
result4.re[,,j]<-apply(cbind(DE=object$ade4[,j],object$aie4[,,j]),2,summary.med.re,object$ate4[,j])}}
else{
result1<-array(0,c(7,2+ncol(object$aie1),ncol(object$ade1),dim(object$ade1)[3]))
result1.re<-array(0,c(7,1+ncol(object$aie1),ncol(object$ade1),dim(object$ade1)[3]))
result2<-result1
result2.re<-result1.re
result3<-result1
result3.re<-result1.re
result4<-result1
result4.re<-result1.re
for(j in 1:ncol(object$ade1)){
for(q1 in 1:dim(object$ade1)[3]){
result1[,,j,q1]<-apply(cbind(TE=object$ate1[,j,q1],DE=object$ade1[,j,q1],object$aie1[,,j,q1]),2,summary.med)
result1.re[,,j,q1]<-apply(cbind(DE=object$ade1[,j,q1],object$aie1[,,j,q1]),2,summary.med.re,object$ate1[,j,q1])
result2[,,j,q1]<-apply(cbind(TE=object$ate2[,j,q1],DE=object$ade2[,j,q1],object$aie2[,,j,q1]),2,summary.med)
result2.re[,,j,q1]<-apply(cbind(DE=object$ade2[,j,q1],object$aie2[,,j,q1]),2,summary.med.re,object$ate2[,j,q1])
result3[,,j,q1]<-apply(cbind(TE=object$ate3[,j,q1],DE=object$ade3[,j,q1],object$aie3[,,j,q1]),2,summary.med)
result3.re[,,j,q1]<-apply(cbind(DE=object$ade3[,j,q1],object$aie3[,,j,q1]),2,summary.med.re,object$ate3[,j,q1])
result4[,,j,q1]<-apply(cbind(TE=object$ate4[,j,q1],DE=object$ade4[,j,q1],object$aie4[,,j,q1]),2,summary.med)
result4.re[,,j,q1]<-apply(cbind(DE=object$ade4[,j,q1],object$aie4[,,j,q1]),2,summary.med.re,object$ate4[,j,q1])}
}
}
c.names=colnames(object$aie1)
colnames(result1)=c("TE","DE",c.names)
colnames(result2)=c("TE","DE",c.names)
colnames(result3)=c("TE","DE",c.names)
colnames(result4)=c("TE","DE",c.names)
colnames(result1.re)=c("DE",c.names)
colnames(result2.re)=c("DE",c.names)
colnames(result3.re)=c("DE",c.names)
colnames(result4.re)=c("DE",c.names)
rownames(result1)=c("mean","sd",paste("q",quant,sep="_"))
rownames(result2)=c("mean","sd",paste("q",quant,sep="_"))
rownames(result3)=c("mean","sd",paste("q",quant,sep="_"))
rownames(result4)=c("mean","sd",paste("q",quant,sep="_"))
rownames(result1.re)=c("mean","sd",paste("q",quant,sep="_"))
rownames(result2.re)=c("mean","sd",paste("q",quant,sep="_"))
rownames(result3.re)=c("mean","sd",paste("q",quant,sep="_"))
rownames(result4.re)=c("mean","sd",paste("q",quant,sep="_"))
result=list(result1=result1,result1.re=result1.re,result2=result2,result2.re=result2.re,
result3=result3,result3.re=result3.re,result4=result4,result4.re=result4.re,method=method,
digit=digit,plot=plot,RE=RE,y.type=object$data0$y_type)
class(result)="summary.bma"
result
}
print.summary.bma<-function (x, ..., digit = x$digit, method=x$method, RE=x$RE)
{plot.sum<-function(obj1,main1="Estimated Effects")
{oldpar <- par(no.readonly = TRUE)
on.exit(par(oldpar))
re <- obj1[1,]
upper <- obj1[7,]
lower <- obj1[3,]
name1 <- colnames(obj1)
par(mfrow = c(1, 1), mar = c(1, 6, 1, 1), oma = c(3, 2, 2, 4))
bp <- barplot2(re, horiz = TRUE, main = main1,
names.arg = name1, plot.ci = TRUE, ci.u = upper,
ci.l = lower, cex.names = 0.9, beside = FALSE,
cex.axis = 0.9, las = 1, xlim = range(c(upper,lower), na.rm = TRUE,finite=T),
col = rainbow(length(re), start = 3/6, end = 4/6))}
if(x$y.type!=3){
for (j in 1:dim(x$result1)[3])
{cat('\n For Predictor', j, ':\n')
if (method==1)
{if(!RE)
{cat('Estimated Effects for Method 1:\n')
print(round(x$result1[,,j],digits = digit))
if(x$plot)
plot.sum(x$result1[,,j],main1=paste("Estimated Effects Using Method", method, "(predictor",j,")"))}
else{
cat('Estimated Relative Effects for Method 1:\n')
print(round(x$result1.re[,,j],digits = digit))
if(x$plot)
plot.sum(x$result1.re[,,j],main1=paste("Estimated Relative Effects Using Method", method, "(predictor",j,")"))}}
else if(method==2){
if(!RE)
{cat('Estimated Effects for Method 2:\n')
print(round(x$result2[,,j],digits = digit))
if(x$plot)
plot.sum(x$result2[,,j],main1=paste("Estimated Effects Using Method", method, "(predictor",j,")"))}
else{
cat('Estimated Relative Effects for Method 2:\n')
print(round(x$result2.re[,,j],digits = digit))
if(x$plot)
plot.sum(x$result2.re[,,j],main1=paste("Estimated Relative Effects Using Method", method, "(predictor",j,")"))}
}
else if(method==3){
if(!RE)
{cat('Estimated Effects for Method 3:\n')
print(round(x$result3[,,j],digits = digit))
if(x$plot)
plot.sum(x$result3[,,j],main1=paste("Estimated Effects Using Method", method, "(predictor",j,")"))}
else{
cat('Estimated Relative Effects for Method 3:\n')
print(round(x$result3.re[,,j],digits = digit))
if(x$plot)
plot.sum(x$result3.re[,,j],main1=paste("Estimated Relative Effects Using Method", method, "(predictor",j,")"))}
}
else if(method==4){
if(!RE)
{cat('Estimated Effects for Method 4:\n')
print(round(x$result4[,,j],digits = digit))
if(x$plot)
plot.sum(x$result4[,,j],main1=paste("Estimated Effects Using Method", method, "(predictor",j,")"))}
else{
cat('Estimated Relative Effects for Method 4:\n')
print(round(x$result4.re[,,j],digits = digit))
if(x$plot)
plot.sum(x$result4.re[,,j],main1=paste("Estimated Relative Effects Using Method", method, "(predictor",j,")"))}
}
}}
else{
for (j in 1:dim(x$result1)[3])
for (q1 in 1:dim(x$result1)[4])
{cat('\n For Predictor', j, 'outcome',q1, ':\n')
if (method==1)
{if(!RE)
{cat('Estimated Effects for Method 1:\n')
print(round(x$result1[,,j,q1],digits = digit))
if(x$plot)
plot.sum(x$result1[,,j,q1],main1=paste("Estimated Effects Using Method", method, "(predictor",j,"outcome",q1,")"))}
else{
cat('Estimated Relative Effects for Method 1:\n')
print(round(x$result1.re[,,j,q1],digits = digit))
if(x$plot)
plot.sum(x$result1.re[,,j,q1],main1=paste("Estimated Relative Effects Using Method", method, "(predictor",j,"outcome",q1,")"))}}
else if(method==2){
if(!RE)
{cat('Estimated Effects for Method 2:\n')
print(round(x$result2[,,j,q1],digits = digit))
if(x$plot)
plot.sum(x$result2[,,j,q1],main1=paste("Estimated Effects Using Method", method, "(predictor",j,"outcome",q1,")"))}
else{
cat('Estimated Relative Effects for Method 2:\n')
print(round(x$result2.re[,,j,q1],digits = digit))
if(x$plot)
plot.sum(x$result2.re[,,j,q1],main1=paste("Estimated Relative Effects Using Method", method, "(predictor",j,"outcome",q1,")"))}
}
else if(method==3){
if(!RE)
{cat('Estimated Effects for Method 3:\n')
print(round(x$result3[,,j,q1],digits = digit))
if(x$plot)
plot.sum(x$result3[,,j,q1],main1=paste("Estimated Effects Using Method", method, "(predictor",j,"outcome",q1,")"))}
else{
cat('Estimated Relative Effects for Method 3:\n')
print(round(x$result3.re[,,j,q1],digits = digit))
if(x$plot)
plot.sum(x$result3.re[,,j,q1],main1=paste("Estimated Relative Effects Using Method", method, "(predictor",j,"outcome",q1,")"))}
}
else if(method==4){
if(!RE)
{cat('Estimated Effects for Method 4:\n')
print(round(x$result4[,,j,q1],digits = digit))
if(x$plot)
plot.sum(x$result4[,,j,q1],main1=paste("Estimated Effects Using Method", method, "(predictor",j,"outcome",q1,")"))}
else{
cat('Estimated Relative Effects for Method 4:\n')
print(round(x$result4.re[,,j,q1],digits = digit))
if(x$plot)
plot.sum(x$result4.re[,,j,q1],main1=paste("Estimated Relative Effects Using Method", method, "(predictor",j,"outcome",q1,")"))}
}
}
}
}
Try the BayesianMediationA package in your browser
Any scripts or data that you put into this service are public.
BayesianMediationA documentation built on Sept. 25, 2022, 1:05 a.m.
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Defines functions summary.bma.bx.cy bma.bx.cy data_org
Documented in bma.bx.cy data_org summary.bma.bx.cy
data_org<-function(pred, m, y, refy = rep(NA, ncol(data.frame(y))),
predref = rep(NA, ncol(data.frame(pred))), fpy = NULL,
deltap = rep(0.001,ncol(data.frame(pred))), fmy = NULL,
deltam = rep(0.001,ncol(data.frame(m))), fpm = NULL,
mref = rep(NA, ncol(data.frame(m))),cova = NULL,
mcov = NULL, mclist=NULL) #mcov is the data frame with all covariates for TVs, mind is the indicator for covariates
#if mclist is null but not mcov, mcov is applied to all tvs.
{ns.dev<-function (x, df = NULL, knots = NULL, qnots=NULL,intercept = FALSE, Boundary.knots = range(x),derivs1=0)
{
nx <- names(x)
x <- as.vector(x)
nax <- is.na(x)
if (nas <- any(nax))
x <- x[!nax]
if (!missing(Boundary.knots)) {
Boundary.knots <- sort(Boundary.knots)
outside <- (ol <- x < Boundary.knots[1L]) | (or <- x >
Boundary.knots[2L])
}
else outside <- FALSE
if (!is.null(df) && is.null(knots)) {
nIknots <- df - 1L - intercept
if (nIknots < 0L) {
nIknots <- 0L
warning(gettextf("'df' was too small; have used %d",
1L + intercept), domain = NA)
}
knots <- if (nIknots > 0L) {
knots <- seq.int(0, 1, length.out = nIknots + 2L)[-c(1L,
nIknots + 2L)]
stats::quantile(x[!outside], knots)
}
}
else {if(is.null(df) && is.null(knots) && !is.null(qnots))
knots<-quantile(x[!outside], qnots)
nIknots <- length(knots)}
Aknots <- sort(c(rep(Boundary.knots, 4L), knots))
if (any(outside)) {
basis <- array(0, c(length(x), nIknots + 4L))
if (any(ol)) {
k.pivot <- Boundary.knots[1L]
xl <- cbind(1, x[ol] - k.pivot)
tt <- splineDesign(Aknots, rep(k.pivot, 2L), 4, c(0,
1),derivs=rep(derivs1,2L))
basis[ol, ] <- xl %*% tt
}
if (any(or)) {
k.pivot <- Boundary.knots[2L]
xr <- cbind(1, x[or] - k.pivot)
tt <- splineDesign(Aknots, rep(k.pivot, 2L), 4, c(0,1),derivs=rep(derivs1,2L))
basis[or, ] <- xr %*% tt
}
if (any(inside <- !outside))
basis[inside, ] <- splineDesign(Aknots, x[inside],
4,derivs=rep(derivs1,length(x[inside])))
}
else basis <- splineDesign(Aknots, x, 4,derivs=rep(derivs1,length(x)))
const <- splineDesign(Aknots, Boundary.knots, 4, c(2, 2),derivs=rep(derivs1,length(Boundary.knots)))
if (!intercept) {
const <- const[, -1, drop = FALSE]
basis <- basis[, -1, drop = FALSE]
}
qr.const <- qr(t(const))
basis <- as.matrix((t(qr.qty(qr.const, t(basis))))[, -(1L:2L),
drop = FALSE])
n.col <- ncol(basis)
if (nas) {
nmat <- matrix(NA, length(nax), n.col)
nmat[!nax, ] <- basis
basis <- nmat
}
dimnames(basis) <- list(nx, 1L:n.col)
a <- list(degree = 3L, knots = if (is.null(knots)) numeric() else knots,
Boundary.knots = Boundary.knots, intercept = intercept)
attributes(basis) <- c(attributes(basis), a)
class(basis) <- c("ns", "basis", "matrix")
basis
}
bs.dev<-function (x, df = NULL, knots = NULL, degree = 3, intercept = FALSE,
Boundary.knots = range(x),derivs1=0)
{
nx <- names(x)
x <- as.vector(x)
nax <- is.na(x)
if (nas <- any(nax))
x <- x[!nax]
if (!missing(Boundary.knots)) {
Boundary.knots <- sort(Boundary.knots)
outside <- (ol <- x < Boundary.knots[1L]) | (or <- x >
Boundary.knots[2L])
}
else outside <- FALSE
ord <- 1L + (degree <- as.integer(degree))
if (ord <= 1)
stop("'degree' must be integer >= 1")
if (!is.null(df) && is.null(knots)) {
nIknots <- df - ord + (1L - intercept)
if (nIknots < 0L) {
nIknots <- 0L
warning(gettextf("'df' was too small; have used %d",
ord - (1L - intercept)), domain = NA)
}
knots <- if (nIknots > 0L) {
knots <- seq.int(from = 0, to = 1, length.out = nIknots +
2L)[-c(1L, nIknots + 2L)]
stats::quantile(x[!outside], knots)
}
}
Aknots <- sort(c(rep(Boundary.knots, ord), knots))
if (any(outside)) {
warning("some 'x' values beyond boundary knots may cause ill-conditioned bases")
derivs <- 0:degree
scalef <- gamma(1L:ord)
basis <- array(0, c(length(x), length(Aknots) - degree -
1L))
if (any(ol)) {
k.pivot <- Boundary.knots[1L]
xl <- cbind(1, outer(x[ol] - k.pivot, 1L:degree,
"^"))
tt <- splineDesign(Aknots, rep(k.pivot, ord), ord,
derivs+derivs1)
basis[ol, ] <- xl %*% (tt/scalef)
}
if (any(or)) {
k.pivot <- Boundary.knots[2L]
xr <- cbind(1, outer(x[or] - k.pivot, 1L:degree,
"^"))
tt <- splineDesign(Aknots, rep(k.pivot, ord), ord,
derivs+derivs1)
basis[or, ] <- xr %*% (tt/scalef)
}
if (any(inside <- !outside))
basis[inside, ] <- splineDesign(Aknots, x[inside],
ord,derivs=rep(derivs1,length(x[inside])))
}
else basis <- splineDesign(Aknots, x, ord, derivs=rep(derivs1,length(x)))
if (!intercept)
basis <- basis[, -1L, drop = FALSE]
n.col <- ncol(basis)
if (nas) {
nmat <- matrix(NA, length(nax), n.col)
nmat[!nax, ] <- basis
basis <- nmat
}
dimnames(basis) <- list(nx, 1L:n.col)
a <- list(degree = degree, knots = if (is.null(knots)) numeric(0L) else knots,
Boundary.knots = Boundary.knots, intercept = intercept)
attributes(basis) <- c(attributes(basis), a)
class(basis) <- c("bs", "basis", "matrix")
basis
}
x2fx<-function(x,func) #x is the list of original numerical vector, func is a vector of character functions.
{ # eg. func <- c("x","x+1","x+2","x+3","log(x)")
func.list <- list()
#test.data <- matrix(data=rep(x,length(func)),length(x),length(func))
#test.data <- data.frame(test.data)
result<-NULL
for(i in 1:length(func)){
func.list[[i]] <- function(x){}
body(func.list[[i]]) <- parse(text=func[i])
}
#result <- mapply(do.call,func.list,lapply(test.data,list))
col_fun<-NULL
z<-1
for (i in 1:length(func.list))
{res<-as.matrix(func.list[[i]](x))
result<-cbind(result,res)
col_fun<-cbind(col_fun,c(z,z+ncol(res)-1))
z<-z+ncol(res)}
list(values=as.matrix(result),col_fun=as.matrix(col_fun))
}
x2fdx<-function(x,func) #x is the list of original numerical vector, func is a vector of character functions.
{ fdx<-NULL # eg. func <- c("x","x+1","x+2","x+3","log(x)")
for(i in 1:length(func)){
if (length(grep("ifelse",func[i]))>0)
{str<-unlist(strsplit(func[i],","))
fun1<-D(parse(text=str[2]), "x")
fun2<-D(parse(text=unlist(strsplit(str[3],")"))),"x")
x1<-eval(fun1)
x2<-eval(fun2)
if(length(x1)==1)
x1<-rep(x1,length(x))
if(length(x2)==1)
x2<-rep(x2,length(x))
fun3<-paste(str[1],"x1,x2)",sep=",")
fdx<-cbind(fdx,eval(parse(text=fun3)))
}
else if(length(grep("ns",func[i]))>0)
{temp<-paste("ns.dev",substring(func[i],3,nchar(func[i])-1),",derivs1=1)",sep="")
fdx<-cbind(fdx,eval(parse(text=temp)))}
else if(length(grep("bs",func[i]))>0)
{temp<-paste("bs.dev",substring(func[i],3,nchar(func[i])-1),",derivs1=1)",sep="")
fdx<-cbind(fdx,eval(parse(text=temp)))}
else{
dx2x <- D(parse(text=func[i]), "x")
temp<-eval(dx2x)
if(length(temp)==1)
fdx<-cbind(fdx,rep(temp,length(x)))
else fdx<-cbind(fdx,temp)}
}
as.matrix(fdx)
}
bin_cat <- function(M2, M1, cat1, catref = 1)
#turn a categorical variable to binary dummy variables
#M2 is the original data frame
#cat1 is the column number of the categorical variable in M2
#catref is the reference group
{a <- factor(M2[, cat1])
b <- sort(unique(a[a != catref]))
d<-NULL
e<-rep(1,nrow(M2))
for(i in 1:length(b))
{d=cbind(d,ifelse(a==b[i],1,0))
e=ifelse(a==b[i],i+1,e)
}
d[is.na(M2[,cat1]),]=NA
e[is.na(M2[,cat1])]=NA
M2[,cat1]=e
xnames=colnames(M1)
M1=cbind(M1,d)
colnames(M1)=c(xnames,paste(colnames(M2)[cat1],b,sep="."))
list(M1=M1,M2=M2,cat=c(ncol(M1)-ncol(d)+1,ncol(M1)))
}
order_char<-function(char1,char2) #find the position of char2 in char1
{a<-1:length(char1)
b<-NULL
for (i in 1:length(char2))
b<-c(b,a[char1==char2[i]])
b
}
###start the main code
#clean up the outcomes:y_type=type of outcomes;
#y_type=2:binary, 3:category, 1:continuous, 4:time-to-event
#consider only 1 outcome for now
#consider only complete case
data.temp=cbind(pred,m,y)
if (!is.null(mcov) & !is.null(mclist))
data.temp=cbind(pred,m,y,mcov)
if(!is.null(cova))
data.temp=cbind(data.temp,cova)
choose.temp=complete.cases(data.temp)
if(ncol(data.frame(pred))==1)
pred=pred[choose.temp]
else
pred=pred[choose.temp,]
if(ncol(data.frame(y))==1)
y=y[choose.temp]
else
y=y[choose.temp,]
if(ncol(data.frame(m))==1)
m=m[choose.temp]
else
m=m[choose.temp,]
if(!is.null(cova)){
if(ncol(data.frame(cova))==1)
cova=cova[choose.temp]
else
cova=cova[choose.temp,]}
if(!is.null(mcov)){
if(ncol(data.frame(mcov))==1)
mcov=mcov[choose.temp]
else
mcov=mcov[choose.temp,]}
if (!is(y,"Surv"))
{if (nlevels(droplevels(as.factor(y))) == 2) {#binary
y_type <- 2
if (!is.na(refy))
y <- ifelse(y == refy, 0, 1)
else {
refy <- levels(droplevels(as.factor(y)))[1]
y <- ifelse(as.factor(y) == refy,0, 1)
}
}
else if (is.character(y) | is.factor(y)) {#categorical
y_type <- 3
y <- droplevels(y)
if (is.na(refy))
refy <- levels(as.factor(y))[1]
a <- factor(y)
b <- sort(unique(a[a != refy]))
e <- rep(1,nrow(as.matrix(y)))
for(j in 1:length(b))
e=ifelse(a==b[j],j+1,e)
e[is.na(y)]=NA
y=factor(e)
}
else #continuous
y_type = 1}
else
{y_type<-4
y=cbind(y[,1],y[,2])}
#clean up m and pred
mnames <- colnames(m)
if (!is.null(cova)) {
if (is.null(colnames(cova)))
cova_names = paste("cova",1:ncol(data.frame(cova)))
else
cova_names = colnames(cova)
cova=data.frame(cova)
colnames(cova)=cova_names
}
if (!is.null(mcov)) {
if (length(grep("for.m", names(mcov))) == 0)
mcov_names = colnames(mcov)
else mcov_names = colnames(mcov[[1]])
mcov=data.frame(mcov)
}
pred_names = names(pred)
##prepare for the predictor(s)
pred1 <- data.frame(pred) #original format
pred1.0 <- NULL #original format with binarized categorical predictor
pred1.0_names <-NULL
pred2 <- NULL #all transformed
pred2_names = NULL
pred3 <- NULL #transformed continuous predictors with pred+delta(pred)
pred3_names = NULL
pred.cont.der <- NULL #derivative of the transformation function for cont predictors
if (is.null(pred_names))
pred_names = paste("pred",1:ncol(pred1),sep='')
colnames(pred1) = pred_names
binpred = NULL #binary predictor in pred2
catpred = NULL #categorical predictor in pred2, each row is for one categorical predictor
contpred = NULL #continuos predictor in pred2, each row is for a continuous predictor
binpred1 = NULL #binary predictor in pred1
catpred1 = NULL #categorical predictor in pred1
contpred1 = NULL #continuous predictor in pred1
binpred1.0 = NULL #binary predictor in pred1.0
catpred1.0 = NULL #categorical predictor in pred1.0
contpred1.0 = NULL #continuous predictor in pred1.0
contpred3 = NULL #index for pred3 and pred.cont.dev
npred = ncol(pred1)
n1=nrow(pred1)
for (i in 1:npred)
if (nlevels(droplevels(as.factor(pred1[,i]))) == 2) { #binary predictor
if (!is.na(predref[i]))
{pred2 <- cbind(pred2,ifelse(pred1[, i] == predref[i],0, 1))
pred1.0 <- cbind(pred1.0, ifelse(pred1[, i] == predref[i],0, 1))
#pred3 <- cbind(pred3,as.factor(ifelse(pred1[, i] == predref[i],0, 1)))
pred1[,i] <- ifelse(pred1[, i] == predref[i],0, 1)}
else {
temp.pred <- as.factor(pred1[, i])
pred2 <- cbind(pred2,ifelse(temp.pred == levels(droplevels(temp.pred))[1], 0, 1))
pred1.0 <- cbind(pred1.0,ifelse(temp.pred == levels(droplevels(temp.pred))[1], 0, 1))
#pred3 <- cbind(pred3,as.factor(ifelse(temp.pred == levels(droplevels(temp.pred))[1], 0, 1)))
pred1[,i] <- ifelse(temp.pred == levels(droplevels(temp.pred))[1], 0, 1)
}
binpred1 = c(binpred1, i)
pred2_names=c(pred2_names,pred_names[i])
pred1.0_names=c(pred1.0_names,pred_names[i])
binpred = c(binpred,ncol(pred2))
binpred1.0 = c(binpred1.0,ncol(pred1.0))
}
else if (is.character(pred1[, i]) | is.factor(pred1[, i])) { #category predictor
pred1[, i] = droplevels(pred1[, i])
if(!is.null(pred2))
colnames(pred2)=pred2_names
if(!is.null(pred1.0))
colnames(pred1.0)=pred1.0_names
# catn = catn + 1
if (!is.na(predref[i]))
{pred.temp1 <- bin_cat(pred1, pred2, i, predref[i])
pred.temp2 <- bin_cat(pred1, pred1.0, i, predref[i])}
else
{pred.temp1 <- bin_cat(pred1, pred2, i, levels(as.factor(pred1[,i]))[1])
pred.temp2 <- bin_cat(pred1, pred1.0, i, levels(as.factor(pred1[,i]))[1])}
pred2 = pred.temp1$M1
pred1.0 = pred.temp2$M1
pred1 = pred.temp1$M2
#pred3 = cbind(pred3,pred.temp1$M1[,pred.temp1$cat[1]:pred.temp1$cat[2]])
catpred1 = c(catpred1,i)
catpred = rbind(catpred,pred.temp1$cat)
catpred1.0 = rbind(catpred1.0,pred.temp2$cat)
pred2_names = colnames(pred.temp1$M1)
pred1.0_names = colnames(pred.temp2$M1)
}
else #consider the transformation of continuous x
{contpred1 = c(contpred1, i)
pred1.0=cbind(pred1.0,pred1[,i])
pred1.0_names<-c(pred1.0_names,pred_names[i])
contpred1.0= c(contpred1.0, ncol(pred1.0))
if(!(i %in% fpy[[1]])) #fpy has the transformation functions for pred to y, [[1]] list all cont pred to be transformed
{pred2<-cbind(pred2,pred1[,i])
pred.cont.der<-cbind(pred.cont.der,rep(1,n1))
contpred3 = rbind(contpred3,c(ncol(pred.cont.der),ncol(pred.cont.der)))
pred3 = cbind(pred3,pred1[,i]+deltap[i]) #deltap is the changing amount for predictors
contpred = rbind(contpred,c(ncol(pred2),ncol(pred2)))
pred2_names=c(pred2_names,pred_names[i])
pred3_names=c(pred3_names,pred_names[i])
}
else if (i %in% fpy[[1]])
{p=match(i,fpy[[1]])
temp.pred=x2fx(pred1[,i],fpy[[1+p]])$values
pred2<-cbind(pred2,temp.pred)
pred.cont.der<-cbind(pred.cont.der,x2fdx(pred1[,i],fpy[[1+p]]))
if(is.null(pred3))
contpred3 = rbind(contpred3,c(1,ncol(pred.cont.der)))
else
contpred3 = rbind(contpred3,c(ncol(pred3)+1,ncol(pred.cont.der)))
pred3<-cbind(pred3,x2fx(pred1[,i]+deltap[i],fpy[[1+p]])$values)
l=length(fpy[[1+p]])
contpred=rbind(contpred,c(ncol(pred2)-ncol(temp.pred)+1,ncol(pred2)))
pred2_names<-c(pred2_names,paste(pred_names[i],1:l,sep="."))
pred3_names=c(pred3_names,paste(pred_names[i],1:l,sep="."))
}
}
colnames(pred1.0)=pred1.0_names
colnames(pred2)=pred2_names
if(!is.null(pred3))
{colnames(pred3)=pred3_names
colnames(pred.cont.der)=pred3_names}
## prepare mediators for y
m1 <- data.frame(m) #original format
m2 <- NULL #all transformed
m2_names = NULL
m3 <- NULL #transformed continuous mediators with mediator+delta(med)
m3_names = NULL
m.cont.der <- NULL #derivative of the transformation function for cont mediators
if (is.null(mnames))
mnames = paste("m",1:ncol(m1),sep='')
colnames(m1) = mnames
binm = NULL
catm = NULL
contm = NULL
binm1 = NULL
catm1 = NULL
contm1 = NULL
contm3 = NULL #index for m3 and m.cont.dev
nm = ncol(m1)
n2=nrow(m1)
for (i in 1:nm)
if (nlevels(droplevels(as.factor(m1[,i]))) == 2) { #binary mediator
if (!is.na(mref[i]))
{m2 <- cbind(m2,ifelse(m1[, i] == mref[i],0, 1))
#m3 <- cbind(m3,as.factor(ifelse(m1[, i] == mref[i],0, 1)))
m1[,i] <- ifelse(m1[, i] == mref[i],0, 1)}
else {
temp.m <- as.factor(m1[, i])
m2 <- cbind(m2,ifelse(temp.m == levels(droplevels(temp.m))[1], 0, 1))
#m3 <- cbind(m3,as.factor(ifelse(temp.m == levels(droplevels(temp.m))[1], 0, 1)))
m1[,i] <- ifelse(temp.m == levels(droplevels(temp.m))[1], 0, 1)
}
binm1 = c(binm1, i)
m2_names=c(m2_names,mnames[i])
colnames(m2)=m2_names
binm = c(binm,ncol(m2))
}
else if (is.character(m1[, i]) | is.factor(m1[, i])) { #category mediator
m1[, i] = droplevels(as.factor(m1[, i]))
if (!is.na(mref[i]))
m.temp1 <- bin_cat(m1, m2, i, mref[i])
else
m.temp1 <- bin_cat(m1, m2, i, levels(as.factor(m1[,i]))[1])
m2 = m.temp1$M1
m1 = m.temp1$M2
#m3 = cbind(m3,m.temp1$M1[,m.temp1$cat[1]:m.temp1$cat[2]])
catm1 = c(catm1,i)
catm = rbind(catm,m.temp1$cat)
m2_names = c(m2_names, colnames(m.temp1$M1)[m.temp1$cat[1]:m.temp1$cat[2]])
colnames(m2)=m2_names
}
else #consider the transformation of continuous m
{contm1 = c(contm1, i)
if(!(i %in% fmy[[1]])) #fmy has the transformation functions for m to y, [[1]] list all cont mediators in m to be transformed
{m2<-cbind(m2,m1[,i])
m.cont.der<-cbind(m.cont.der,rep(1,n1))
contm3 = rbind(contm3,c(ncol(m.cont.der),ncol(m.cont.der)))
m3 = cbind(m3,m1[,i]+deltam[i]) #deltam is the changing amount for mediators
m2_names=c(m2_names,mnames[i])
colnames(m2)=m2_names
m3_names=c(m3_names,mnames[i])
contm=rbind(contm,c(ncol(m2),ncol(m2)))
}
else if (i %in% fmy[[1]])
{p=match(i,fmy[[1]])
temp.m=x2fx(m1[,i],fmy[[1+p]])$values
m2<-cbind(m2,temp.m)
m.cont.der<-cbind(m.cont.der,x2fdx(as.matrix(m)[,i],fmy[[1+p]]))
if(is.null(m3))
contm3 = rbind(contm3,c(1,ncol(m.cont.der)))
else
contm3 = rbind(contm3,c(ncol(m3)+1,ncol(m.cont.der)))
m3<-cbind(m3,x2fx(m1[,i]+deltap[i],fmy[[1+p]])$values)
contm = rbind(contm,c(ncol(m2)-ncol(temp.m)+1,ncol(m2)))
l=length(fmy[[1+p]])
m2_names<-c(m2_names,paste(mnames[i],1:l,sep="."))
colnames(m2)=m2_names
m3_names=c(m3_names,paste(mnames[i],1:l,sep="."))
}
}
colnames(m2)=m2_names
if(!is.null(m3))
{colnames(m3)=m3_names
colnames(m.cont.der)=m3_names}
## prepare predictors for mediators
pm=rep(0,n1) #the first row are all 0s
fpm.2=fpm #the transformed predictors in pm
binp=NULL
catp=NULL
contp=NULL
j=1
names.pm=("zero")
if(!is.null(binpred))
for (i in binpred)
{pm=cbind(pm,pred2[,i])
j=j+1
binp=c(binp,j)
names.pm=c(names.pm,pred_names[i])}
if(!is.null(catpred))
for (i in 1:nrow(catpred))
{pm=cbind(pm,pred2[,catpred[i,1]:catpred[i,2]])
catp=rbind(catp,c(j+1,j+1+catpred[i,2]-catpred[i,1]))
j=j+1+catpred[i,2]-catpred[i,1]
names.pm=c(names.pm,colnames(pred2)[catpred[i,1]:catpred[i,2]])
}
if(!is.null(contpred1))
for (i in contpred1)
{pm=cbind(pm,pred1[,i])
j=j+1
contp=rbind(contp,rep(j,3))
names.pm=c(names.pm,pred_names[i])}
pm.der=matrix(1,n1,ncol(pm))
pm.idx=as.list(rep(1,ncol(m1))) #the predictors in pm to predict the ith mediator
for (i in 1:ncol(m1))
pm.idx[[i]]=2:ncol(pm)
if(!is.null(fpm))
{k=unique(fpm[[1]][,2]) #the first column is for the mediators,
#the second column the continuous predictors to be transformed
for (l in k){
temp<-(2:length(fpm))[fpm[[1]][,2]==l]
allfun=fpm[[temp[1]]]
if (length(temp)>1)
for(i in 2:length(temp))
allfun<-c(allfun,fpm[[temp[i]]])
unifun<-unique(allfun)
unifun1<-unifun[unifun!="x"]
unifun2<-c("x",unifun1)
d_d<-x2fx(pred1[,l],unifun1)
d.der<-x2fdx(pred1[,l],unifun1)
d<-as.matrix(d_d$values)
names.pm<-c(names.pm,paste(pred_names[l],1:ncol(d),sep="."))
pm.der<-cbind(pm.der,d.der)
place=match(l,contpred1)
contp[place,2]=j+1
contp[place,3]=j+ncol(d)
pm<-cbind(pm,d)
for(i in temp)
{ttemp<-order_char(unifun1,fpm[[i]])
fpm.2[[i]]=j+ttemp #what does this do?
pm.indx[[fpm[[1]][i-1,1]]]=c(pm.indx[[fpm[[1]][i-1,1]]],j+ttemp)
if(length(order_char('x',fpm[[i]]))==0)
pm.indx[[fpm[[1]][i-1,1]]]= (pm.indx[[fpm[[1]][i-1,1]]])[pm.indx[[fpm[[1]][i-1,1]]]!=l]
}
j=j+ncol(d)
}}
colnames(pm)=names.pm
colnames(pm.der)=names.pm
pm.ind=matrix(1,length(pm.idx),max(sapply(pm.idx,length)))
for (i in 1:nrow(pm.ind))
pm.ind[i,1:length(pm.idx[[i]])]=pm.idx[[i]]
p2=ifelse(is.null(binm1),0,length(binm1))
p3=ifelse(is.null(catm1),0,length(catm1))
p1=ifelse(is.null(contm1),0,length(contm1))
#prepare for covariates of mediators
if(is.null(mcov))
{mcov=data.frame(intercept=rep(1,nrow(m1)))
mind=matrix(T,p1+p2+p3,1)}
else
{mcov=data.frame(intercept=rep(1,nrow(m)),mcov)
mind=matrix(T,p1+p2+p3,ncol(mcov))
mcov_names=colnames(mcov)
if (!is.null(mclist))
{if (is.character(mclist[[1]]))
mclist[[1]]=match(mclist[[1]],mnames)
mcov=data.frame(mcov,no=rep(0,nrow(mcov))) #add a column of 0 in mcov
mind=matrix(rep(1:(ncol(mcov)-1),each=p1+p2+p3),p1+p2+p3,ncol(mcov)-1)
for (i in 1:length(mclist[[1]]))
{if(sum(is.na(mclist[[i+1]]))>=1)
temp=1
else if(is.character(mclist[[i+1]]))
temp=c(1,match(mclist[[i+1]],mcov_names))
else
temp=c(1,mclist[[i+1]]+1)
mind[mclist[[1]][i],(1:(ncol(mcov)-1))[-temp]]=ncol(mcov)
}}
} #use all covariates for all tv if mclist is NULL. Otherwise, the first item of mclist lists all tvs
#that are using different mcov, the following items gives the mcov for the tv in order. use NA is no mcov to be used
results = list(N=nrow(data.frame(y)), y_type=y_type, y=y, pred1=pred1, pred1.0=pred1.0,
pred2=pred2, pred3=pred3, cova=cova,
pred.cont.der=pred.cont.der, binpred2=binpred, catpred2=catpred,
contpred2=contpred, binpred1=binpred1, catpred1=catpred1,contpred1=contpred1,
contpred1.0=contpred1.0, binpred1.0=binpred1.0, catpred1.0=catpred1.0,
contpred3=contpred3, npred=npred,
m1=m1, m2=m2, m3=m3, m.cont.der=m.cont.der, binm2=binm, catm2=catm,
contm2=contm,binm1=binm1, catm1=catm1, contm1=contm1, contm3=contm3,
nm=nm, pm=pm, pm.der=pm.der, pm.idx=pm.idx, pm.ind=pm.ind, fpm.2=fpm.2,
binp=binp, catp=catp, contp=contp,p1=p1,p2=p2,p3=p3,mcov=mcov,mind=mind)
return(results)
}
bma.bx.cy<-function(pred, m, y, refy = rep(NA, ncol(data.frame(y))),
predref = rep(NA, ncol(data.frame(pred))), fpy = NULL,
deltap = rep(0.001,ncol(data.frame(pred))), fmy = NULL,
deltam = rep(0.001,ncol(data.frame(m))), fpm = NULL,
mref = rep(NA, ncol(data.frame(m))),cova = NULL,
mcov = NULL, mclist=NULL, inits=NULL,
n.chains = 1, n.iter = 1100,n.burnin=100,n.thin = 1,
mu=NULL, Omega=NULL, Omegac=NULL,muc=NULL,mucv=NULL, Omegacv=NULL,
mu0.1=NULL, Omega0.1=NULL, mu1.1=NULL, Omega1.1=NULL,
mu0.a=NULL, Omega0.a=NULL, mu1.a=NULL, Omega1.a=NULL,
mu0.b=NULL, Omega0.b=NULL, mu1.b=NULL, Omega1.b=NULL,
mu0.c=NULL, Omega0.c=NULL, mu1.c=NULL, Omega1.c=NULL,
preci=0.000001,tmax=Inf,multi=NULL,filename=NULL)
{ ### build the bugs model if it is not defined
jags_model <- function (contm=c('ncontm', 'ycontm'),binm=c('nbinm','ybinm'),
catm=c('ncatm','ycatm'),
cova=c('ncova','ycova'),
ytype=c('contc','binc','catc','survc','contnnc',
'binnc','catnc','survnc')) {
contm <- match.arg(contm)
binm <- match.arg(binm)
catm <- match.arg(catm)
cova <- match.arg(cova)
# raw script
script <-
"model {temp <- 1:nmc
for(i in 1:N){
$yfunc
$contm
$binm
$catm
}
$ypriors
$cntmprior
$bmprior
$cmprior
var4 ~ dgamma(1,0.1)
prec4 <-1/var4
}"
# define macros
macros <- list(list("$yfunc",
switch(ytype,
contc='mu_y[i] <- beta0 + inprod(c, x[i,]) + inprod(beta,M1[i,]) + inprod(eta,cova[i,])
y[i] ~ dnorm(mu_y[i],prec4)',
contnc='mu_y[i] <- beta0 + inprod(c, x[i,]) + inprod(beta,M1[i,])
y[i] ~ dnorm(mu_y[i],prec4)',
binc='logit(mu_y[i]) <- beta0 + inprod(c, x[i,]) + inprod(beta,M1[i,]) + inprod(eta,cova[i,])
y[i] ~ dbern(mu_y[i])',
binnc='logit(mu_y[i]) <- beta0 + inprod(c, x[i,]) + inprod(beta,M1[i,])
y[i] ~ dbern(mu_y[i])',
catc='mu_y1[i,1] <- 1
for (k in 2:caty)
{mu_y1[i,k] <- exp(beta0[k-1] + inprod(c[k-1,], x[i,]) + inprod(beta[k-1,],M1[i,]) + inprod(eta[k-1,],cova[i,]))}
sum_y[i] <- sum(mu_y1[i,1:caty])
for (l in 1:caty)
{mu_y[i,l] <- mu_y1[i,l]/sum_y[i]}
y[i] ~ dcat(mu_y[i,])',
catnc='mu_y1[i,1] <- 1
for (k in 2:caty)
{mu_y1[i,k] <- exp(beta0[k-1] + inprod(c[k-1,], x[i,]) + inprod(beta[k-1,],M1[i,]))}
sum_y[i] <- sum(mu_y1[i,1:caty])
for (l in 1:caty)
{mu_y[i,l] <- mu_y1[i,l]/sum_y[i]}
y[i]~dcat(mu_y[i,])',
survc= 'elinpred[i] <- exp(inprod(c, x[i,]) + inprod(beta,M1[i,]) + inprod(eta,cova[i,]))
base[i] <- lambda*r*pow(y[i,1], r-1)
loghaz[i] <- log(base[i]*elinpred[i])
phi[i] <- 100000-y[i,2]*loghaz[i]-log(exp(-lambda*pow(y[i,1],r)*elinpred[i])-exp(-lambda*pow(tmax,r)*elinpred[i])) +log(1-exp(-lambda*pow(tmax,r)*elinpred[i]))
zero[i] ~ dpois(phi[i])' ,
survnc= 'elinpred[i] <- exp(inprod(c, x[i,]) + inprod(beta,M1[i,]))
base[i] <- lambda*r*pow(y[i,1], r-1)
loghaz[i] <- log(base[i]*elinpred[i])
phi[i] <- 100000-y[i,2]*loghaz[i]-log(exp(-lambda*pow(y[i,1],r)*elinpred[i])-exp(-lambda*pow(tmax,r)*elinpred[i])) +log(1-exp(-lambda*pow(tmax,r)*elinpred[i]))
zero[i] ~ dpois(phi[i])')),
list("$contm",
switch(contm,
ycontm='for (j in 1:p1){
mu_M1[i,contm[j]] <- inprod(alpha0.a[j,mind[contm[j],]],mcov[i,temp[mind[contm[j],]]])+inprod(alpha1.a[j,1:c1],x1[i,])
M2[i,contm[j]] ~ dnorm(mu_M1[i,contm[j]],prec1[j])
for (k in contm1[j,1]:contm1[j,2]){
mu_M1_c[i,k] <- inprod(alpha0[k,mind[contm[j],]],mcov[i,mind[contm[j],]])+inprod(alpha1[k,1:c1],x1[i,])
M1[i,k] ~ dnorm(mu_M1_c[i,k],prec2[k])
}
}
',
ncontm='')),
list("$cntmprior",
switch(contm,
ycontm=' for(j in 1:P)
{alpha1[j,1:c1] ~ dmnorm(mu1.1[j,1:c1], Omega1.1[1:c1, 1:c1])
alpha0[j,1:nmc] ~ dmnorm(mu0.1[j,1:nmc], Omega0.1[1:nmc, 1:nmc])}
for (i in 1:P){
var2[i] ~ dgamma(1,0.1)
prec2[i] <- 1/var2[i]
}
for(j in 1:p1)
{alpha1.a[j,1:c1] ~ dmnorm(mu1.a[j,1:c1], Omega1.a[1:c1, 1:c1])
alpha0.a[j,1:nmc] ~ dmnorm(mu0.a[j,1:nmc], Omega0.a[1:nmc, 1:nmc])}
for (i in 1:p1){
var1[i] ~ dgamma(1,0.1)
prec1[i] <- 1/var1[i]
}',
ncontm='')),
list("$binm",
switch(binm,
ybinm=" for (k in 1:p2){
logit(mu_M1[i,binm[k]]) <- inprod(alpha0.b[k,mind[binm[k],]],mcov[i,mind[binm[k],]])+inprod(alpha1.b[k,1:c1],x1[i,])
M2[i,binm[k]] ~ dbern(mu_M1[i,binm[k]])
}",
nbinm='')),
list("$bmprior",
switch(binm,
nbinm='',
ybinm=' for(j in 1:p2)
{alpha1.b[j,1:c1] ~ dmnorm(mu1.b[j,1:c1], Omega1.b[1:c1, 1:c1])
alpha0.b[j,1:nmc] ~ dmnorm(mu0.b[j,1:nmc], Omega0.b[1:nmc, 1:nmc])
}')),
list("$catm",
switch(catm,
ycatm=" for (j in 1:p3){
mu_Mc[i,j,1] <- 1 #baseline is the 1st category
for (k in 2:cat2[j]){
mu_Mc[i,j,k] <- exp(inprod(alpha0.c[j,k-1,mind[catm[j],]],mcov[i,mind[catm[j],]])+inprod(alpha1.c[j,k-1,1:c1],x1[i,]))
}
sum_Mc[i,j] <- sum(mu_Mc[i,j,1:cat2[j]])
for (l in 1:cat2[j])
{mu_Mc0[i,j,l] <- mu_Mc[i,j,l]/sum_Mc[i,j]}
M2[i,catm[j]] ~ dcat(mu_Mc0[i,j,1:cat2[j]])
}",
ncatm='')),
list("$cmprior",
switch(catm,
ncatm='',
ycatm=' for (i in 1:p3){
for(j in 1:cat1)
{alpha1.c[i,j,1:c1] ~ dmnorm(mu1.c[j,1:c1], Omega1.c[1:c1, 1:c1])
alpha0.c[i,j,1:nmc] ~ dmnorm(mu0.c[j,1:nmc], Omega0.c[1:nmc, 1:nmc])}
}')),
list("$ypriors",
switch(ytype,
contc=' beta[1:P] ~ dmnorm(mu[1:P], Omega[1:P, 1:P])
beta0 ~ dnorm(0, 1.0E-6)
c[1:c2] ~ dmnorm(muc[1:c2], Omegac[1:c2,1:c2])
eta[1:cv1] ~ dmnorm(mucv[1:cv1], Omegacv[1:cv1,1:cv1])',
contnc=' beta[1:P] ~ dmnorm(mu[1:P], Omega[1:P, 1:P])
beta0 ~ dnorm(0, 1.0E-6)
c[1:c2] ~ dmnorm(muc[1:c2], Omegac[1:c2,1:c2])
',
binc=' beta[1:P] ~ dmnorm(mu[1:P], Omega[1:P, 1:P])
beta0 ~ dnorm(0, 1.0E-6)
c[1:c2] ~ dmnorm(muc[1:c2], Omegac[1:c2,1:c2])
eta[1:cv1] ~ dmnorm(mucv[1:cv1], Omegacv[1:cv1,1:cv1])',
binnc=' beta[1:P] ~ dmnorm(mu[1:P], Omega[1:P, 1:P])
beta0 ~ dnorm(0, 1.0E-6)
c[1:c2] ~ dmnorm(muc[1:c2], Omegac[1:c2,1:c2])
',
catc=' for(j in 1:(caty-1))
{beta[j,1:P] ~ dmnorm(mu[1:P], Omega[1:P, 1:P])
beta0[j] ~ dnorm(0, 1.0E-6)
c[j,1:c2] ~ dmnorm(muc[1:c2], Omegac[1:c2,1:c2])
eta[j,1:cv1] ~ dmnorm(mucv[1:cv1], Omegacv[1:cv1,1:cv1])}',
catnc=' for(j in 1:(caty-1))
{beta[j,1:P] ~ dmnorm(mu[1:P], Omega[1:P, 1:P])
beta0[j] ~ dnorm(0, 1.0E-6)
c[j,1:c2] ~ dmnorm(muc[1:c2], Omegac[1:c2,1:c2])}',
survc= ' beta[1:P] ~ dmnorm(mu[1:P], Omega[1:P, 1:P])
beta0 ~ dnorm(0, 1.0E-6)
c[1:c2] ~ dmnorm(muc[1:c2], Omegac[1:c2,1:c2])
r~duif(0,10) # dunif(0.5,1.5)
lambda~dgamma(1,0.01)
eta[1:cv1] ~ dmnorm(mucv[1:cv1], Omegacv[1:cv1,1:cv1])',
survnc= ' beta[1:P] ~ dmnorm(mu[1:P], Omega[1:P, 1:P])
beta0 ~ dnorm(0, 1.0E-6)
c[1:c2] ~ dmnorm(muc[1:c2], Omegac[1:c2,1:c2])
r ~ dunif(0,10) # dunif(1,1.2)
lambda ~ dgamma(1,0.01)')) # dunif(1.0E-8,3.0E-7)
)
# apply macros
for (m in seq(macros)) {
script <- gsub(macros[[m]][1], macros[[m]][2], script, fixed=TRUE)
}
script
}
### Incomplete Gamma function
Igamma<-function(z,u)pgamma(u,z)*gamma(z)
### Truncated Weibull distribution moments
weib.trunc.mean<-function(vec, right)vec[-1]*Igamma(1/vec[1]+1,(right/vec[-1])^vec[1])/(1-exp(-(right/vec[-1])^vec[1]))
matrix.prod<-function(m1)
{return(m1[1,]%*%t(m1[-1,]))}
expp<-function(v1,v2)
{v1^v2}
data0<- data_org(pred=pred, m=m, y=y, refy=refy, predref=predref, fpy=fpy,
deltap=deltap, fmy=fmy, deltam=deltam, fpm=fpm, mref=mref,
cova=cova,mcov = mcov, mclist=mclist)
y.type=data0$y_type #1 for continuous outcome, 4 for time-to-event, 2 for binary, 3 for categorical
N=data0$N
x=data0$pred2
c2=ncol(x)
x1=data0$pred1.0
c1=ncol(x1) #c1 is the number of predictors, k-class categorical variable are considered as k-1 predictors
y=data0$y
M1=data0$m2
M2=data0$m1
M3=data0$m3
contm=data0$contm1
contm1=data0$contm2
contm3=data0$contm3
p1=data0$p1
binm=data0$binm1
binm1=data0$binm2
p2=data0$p2
p3=data0$p3
if(p3>0){
cat1=max(data0$catm2[,2]-data0$catm2[,1]+1)
cat2=data0$catm2[,2]-data0$catm2[,1]+2
catm=data0$catm1
catm1=data0$catm2}
else{
cat1=NULL
cat2=NULL
catm=NULL
catm1=NULL
}
P=ncol(data0$m2)
cova=data0$cova
mcov=data0$mcov
mind=data0$mind
nmc=ncol(mcov)
# pm=data0$pm
# pm.ind=data1$pm.ind
if(is.null(mu))
mu=rep(0,P)
if(is.null(Omega))
Omega=diag(preci,P)
if(p1>0){
if(is.null(mu0.1))
mu0.1=matrix(0,P,nmc)
if(is.null(Omega0.1))
Omega0.1=diag(preci,nmc)
if(is.null(mu1.1))
mu1.1=matrix(0,P,c1)
if(is.null(Omega1.1))
Omega1.1=diag(preci,c1)
if(is.null(mu0.a))
mu0.a=matrix(0,p1,nmc)
if(is.null(Omega0.a))
Omega0.a=diag(preci,nmc)
if(is.null(mu1.a))
mu1.a=matrix(0,p1,c1)
if(is.null(Omega1.a))
Omega1.a=diag(preci,c1)
}
if(p2>0){
if(is.null(mu0.b))
mu0.b=matrix(0,p2,nmc)
if(is.null(Omega0.b))
Omega0.b=diag(preci,nmc)
if(is.null(mu1.b))
mu1.b=matrix(0,p2,c1)
if(is.null(Omega1.b))
Omega1.b=diag(preci,c1)
}
if(p3>0){
if(is.null(mu0.c))
mu0.c=matrix(0,cat1,nmc)
if(is.null(Omega0.c))
Omega0.c=diag(preci,nmc)
if(is.null(mu1.c))
mu1.c=matrix(0,cat1,c1)
if(is.null(Omega1.c))
Omega1.c=diag(preci,c1)
}
if(is.null(muc))
muc=rep(0,c2)
if(is.null(Omegac))
Omegac=diag(preci,c2)
# data0.1<- list (N=N,x=x,y=y,M1=M1,M2=M2,contm=contm,contm1=contm1,p1=p1,
# binm=binm,p2=p2,cat1=cat1,cat2=cat2,catm=catm,p3=p3,P=P,
# mu=mu,Omega=Omega,mu0.1=mu0.1,mu1.1=mu1.1,Omega0.1=Omega0.1,Omega1.1=Omega1.1,
# mu0.a=mu0.a,mu1.a=mu1.a,Omega0.a=Omega0.a,Omega1.a=Omega1.a,
# mu0.b=mu0.b,mu1.b=mu1.b,Omega0.b=Omega0.b,Omega1.b=Omega1.b,
# mu0.c=mu0.c,mu1.c=mu1.c,Omega0.c=Omega0.c,Omega1.c=Omega1.c)
if(y.type==3)
{caty=nlevels(y)}
if(y.type==4)
{zero=rep(0,nrow(y))#is.cen=ifelse(y[,2]==1,0,1) #censored or not
if(is.null(tmax))
tmax=max(y[,1], na.rm=T)+100
if(is.null(multi))
multi=TRUE} #maximum time to event
#Cen=ifelse(y[,2]==1,tmax,y[,1]) #censoring time
#y=ifelse(y[,2]==1,y[,1],NA)}
data0.1<- list (N=N,x=x,x1=x1,y=y,M1=M1,M2=M2,Omega=Omega,mu=mu,P=P,c1=c1,c2=c2,
muc=muc,Omegac=Omegac,nmc=nmc,mcov=mcov,mind=mind)
para=c("beta0","c","beta")
if(y.type==3)
{data0.1[['caty']]=caty}
if(y.type==4)
{data0.1[['zero']]=zero
#data0.1[['is.cen']]=is.cen
data0.1[['tmax']]=tmax
para=c(para,"lambda","r")}
# if(c1>1)
# {data0.1[['muc']]=rep(0,c1)
# data0.1[['Omegac']]=diag(0.000001,c1)
# } #data0.1[['c1']]=c1
if(p1>0)
{data0.1[['contm']]=contm
data0.1[['contm1']]=contm1
data0.1[['p1']]=p1
data0.1[['mu0.1']]=mu0.1
data0.1[['mu1.1']]=mu1.1
data0.1[['Omega0.1']]=Omega0.1
data0.1[['Omega1.1']]=Omega1.1
data0.1[['mu0.a']]=mu0.a
data0.1[['mu1.a']]=mu1.a
data0.1[['Omega0.a']]=Omega0.a
data0.1[['Omega1.a']]=Omega1.a
para=c(para,"alpha0","alpha1","alpha0.a","alpha1.a")
}
if(p2>0)
{data0.1[['binm']]=binm
data0.1[['p2']]=p2
data0.1[['mu0.b']]=mu0.b
data0.1[['mu1.b']]=mu1.b
data0.1[['Omega0.b']]=Omega0.b
data0.1[['Omega1.b']]=Omega1.b
para=c(para,"alpha0.b","alpha1.b")}
if(p3>0)
{data0.1[['cat1']]=cat1
data0.1[['cat2']]=cat2
data0.1[['catm']]=catm
data0.1[['p3']]=p3
data0.1[['mu0.c']]=mu0.c
data0.1[['mu1.c']]=mu1.c
data0.1[['Omega0.c']]=Omega0.c
data0.1[['Omega1.c']]=Omega1.c
para=c(para,"alpha0.c","alpha1.c")}
if(!is.null(cova)){
cv1=ncol(cova)
if(is.null(mucv))
mucv=rep(0,cv1)
if(is.null(Omegacv))
Omegacv=diag(preci,cv1)
data0.1[['cv1']]=cv1
data0.1[['mucv']]=mucv
data0.1[['Omegacv']]=Omegacv
data0.1[['cova']]=cova
para=c(para,"eta")
}
if(is.null(inits))
inits<- function(){list()}
if (y.type==1 & is.null(cova))
ytype="contnc"
else if(y.type==1 & !is.null(cova))
ytype="contc"
else if(y.type==2 & is.null(cova))
ytype="binnc"
else if(y.type==2 & !is.null(cova))
ytype="binc"
else if(y.type==3 & is.null(cova))
ytype="catnc"
else if(y.type==3 & !is.null(cova))
ytype="catc"
else if(y.type==4 & is.null(cova))
ytype="survnc"
else if(y.type==4 & !is.null(cova))
ytype="survc"
if (is.null(filename))
{filename=paste(tempdir(),"model.txt",sep="/")
writeLines(jags_model(contm=ifelse(p1==0,'ncontm', 'ycontm'),
binm=ifelse(p2==0, 'nbinm','ybinm'),
catm=ifelse(p3==0, 'ncatm','ycatm'),
cova=ifelse(is.null(cova),'ncova','ycova'),
ytype=ytype), filename)
}
med0<- jags(data0.1, inits,
model.file = filename, #"O:/My Documents/My Research/Research/Bayesian Mediation Analysis/codes/promis/bx_cy.txt",
parameters.to.save = para,
n.chains = n.chains, n.iter = n.iter, n.burnin=n.burnin, n.thin = n.thin)
#check the results
#calculate the mediation effects
N1=(n.iter-n.burnin)/n.thin #10000
#m.mcov=apply(mcov,2,mean,na.rm=T) #effects are calculated at the mean of mcov
#med0$BUGSoutput$sims.list$r=1/med0$BUGSoutput$sims.list$r
#attach(med0$BUGSoutput$sims.list)
if(y.type==3){
aie1=array(0,c(N1,p1+p2+p3,c1,caty-1))
ie1=array(0,dim=c(N1,N,p1+p2+p3,c1,caty-1))
tt2=1
de1=array(0,c(N1,c1,caty-1))
n.cont=1
aie2=array(0,c(N1,p1+p2+p3,c1,caty-1))
ie2=array(0,dim=c(N1,N,p1+p2+p3,c1,caty-1))
de2=array(0,c(N1,c1,caty-1))
te3=array(0,c(N1,N,caty-1))
ate3=array(0,c(N1,c1,caty-1))
omu3=ate3
ade3=ate3
aie3=array(0,c(N1,p1+p2+p3,c1,caty-1))
ie3=array(0,dim=c(N1,N,p1+p2+p3,c1,caty-1))
de3=array(0,c(N1,N,caty-1))
mu_M2=array(0,c(dim(M1),N1))
mu_M3=mu_M2
te4=array(0,c(N1,N,caty-1))
de4=te4
ade4=array(0,c(N1,c1,caty-1))
ie4=array(0,dim=c(N1,N,p1+p2+p3,c1,caty-1))
mu.M0=array(0,c(dim(M1),N1))
mu.M1=mu.M0
ate4=ade4
omu4=ate4
aie4=array(0,c(N1,p1+p2+p3,c1,caty-1))
for (l in 1:c1){
x1.temp=x1
x3.temp=x1
if(l%in%data0$contpred1.0){
x3.temp[,l]=x1[,l]+deltap[l]
}
else if(l%in%data0$binpred1.0)
{x1.temp[,l]=0
x3.temp[,l]=1
deltap[l]=1}
else{ #categorical predictor
for (i in 1:nrow(data0$catpred1.0))
if(l%in%(data0$catpred1.0[i,1]:data0$catpred1.0[i,2]))
{x1.temp[,data0$catpred1.0[i,1]:data0$catpred1.0[i,2]]=0
x3.temp[,data0$catpred1.0[i,1]:data0$catpred1.0[i,2]]=0
x3.temp[,l]=1}
deltap[l]=1
}
#method 1: the same for binary or continuous predictors
if(p1>0){
for(k in 1:(caty-1))
if (p1+p2+p3==1 & c1==1 & contm1[1,1]==contm1[1,2])
aie1[,contm[1],l,k]=med0$BUGSoutput$sims.list$alpha1[,1]*med0$BUGSoutput$sims.list$beta[,k,contm1[1,1]]#*mean(data0$m.cont.der[,contm3[1,1]]) since alpha1 is the coefficient of x on the change of f(M), beta is the coefficient of f(M) on the change of y
else
for (j in 1: p1)
aie1[,contm[j],l,k]=apply(as.matrix(med0$BUGSoutput$sims.list$alpha1[,contm1[j,1]:contm1[j,2],l])*as.matrix(med0$BUGSoutput$sims.list$beta[,k,contm1[j,1]:contm1[j,2]]),1,sum)#*mean(data0$m.cont.der[,contm3[j,1]])
}
if(p2>0){
if(p2==1 & c1==1) #since c1=1, the predictor can only be binary or continuous
{if (nmc==1)
{temp.x1=med0$BUGSoutput$sims.list$alpha0.b[,1]+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,1])%*%t(x1.temp)
temp.x3=med0$BUGSoutput$sims.list$alpha0.b[,1]+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,1])%*%t(x3.temp)}
else
{temp.x1=med0$BUGSoutput$sims.list$alpha0.b[,1,mind[binm[1],]]%*%t(mcov[,mind[binm[1],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,1])%*%t(x1.temp)
temp.x3=med0$BUGSoutput$sims.list$alpha0.b[,1,mind[binm[1],]]%*%t(mcov[,mind[binm[1],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,1])%*%t(x3.temp)}
for(j in 1:(caty-1)){
if(!1%in%data0$contpred1.0) #if the predictor is binary
ie1[,,binm[1],1,j]=med0$BUGSoutput$sims.list$beta[,j,binm1[1]]*(exp(temp.x3)/(1+exp(temp.x3))-exp(temp.x1)/(1+exp(temp.x1)))
else #if the predictor is continuous
ie1[,,binm[1],1,j]=med0$BUGSoutput$sims.list$alpha1.b[,1]*med0$BUGSoutput$sims.list$beta[,j,binm1[1]]*exp(temp.x1)/(1+exp(temp.x1))^2
aie1[,binm[1],l,j] <-apply(ie1[,,binm[1],1,j],1,mean)}
}
else
for (k in 1:p2){
if (nmc==1 & p2==1)
{temp.x1=med0$BUGSoutput$sims.list$alpha0.b[,k]+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,k,])%*%t(x1.temp)
temp.x3=med0$BUGSoutput$sims.list$alpha0.b[,k]+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,k,])%*%t(x3.temp)
}
else
{temp.x1=med0$BUGSoutput$sims.list$alpha0.b[,k,mind[binm[k],]]%*%t(mcov[,mind[binm[k],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,k,])%*%t(x1.temp)
temp.x3=med0$BUGSoutput$sims.list$alpha0.b[,k,mind[binm[k],]]%*%t(mcov[,mind[binm[k],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,k,])%*%t(x3.temp)
}
for(j in 1:(caty-1)){
if(!l%in%data0$contpred1.0) #if the predictor is binary or categorical
ie1[,,binm[k],l,j]=med0$BUGSoutput$sims.list$beta[,j,binm1[k]]*(exp(temp.x3)/(1+exp(temp.x3))-exp(temp.x1)/(1+exp(temp.x1)))
else if(data0$contpred1.0) #for continuous predictor
ie1[,,binm[1],l,j]=med0$BUGSoutput$sims.list$alpha1.b[,k,l]*med0$BUGSoutput$sims.list$beta[,j,binm1[k]]*exp(temp.x1)/(1+exp(temp.x1))^2
aie1[,binm[k],l,j] <- apply(ie1[,,binm[k],l,j],1,mean)}
}
}
if(p3>0){
for (j in 1:p3){
mu_Mc1<-array(0,c(N1,N,cat2[j]-1))
mu_Mc0<-array(0,c(N1,N,cat2[j]-1))
for (k in 2:cat2[j]){
mu_Mc0[,,k-1] <- exp(as.matrix(med0$BUGSoutput$sims.list$alpha0.c[,j,k-1,mind[catm[j],]])%*%t(mcov[,mind[catm[j],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.c[,j,k-1,])%*%t(x1.temp))
mu_Mc1[,,k-1] <- exp(as.matrix(med0$BUGSoutput$sims.list$alpha0.c[,j,k-1,mind[catm[j],]])%*%t(mcov[,mind[catm[j],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.c[,j,k-1,])%*%t(x3.temp))
}
sum_Mc1 <-apply(mu_Mc1,c(1,2),sum)+1
sum_Mc0 <-apply(mu_Mc0,c(1,2),sum)+1
if(l%in%data0$contpred1.0) #for continuous predictor
{tt.0=rep(0,N1)
for (k in 2:cat2[j])
tt.0=mu_Mc0[,,k-1]/sum_Mc0*med0$BUGSoutput$sims.list$alpha1.c[,j,k-1,l]
for (q1 in 1:(caty-1))
for (k in 2:cat2[j])
ie1[,,catm[j],l,q1]=ie1[,,catm[j],l,q1]+(mu_Mc0[,,k-1]/sum_Mc0)*med0$BUGSoutput$sims.list$beta[,q1,catm1[j,1]+k-2]*(med0$BUGSoutput$sims.list$alpha1.c[,j,k-1,l]-tt.0)
}
else #for binary or categorical predictor
for (q1 in 1:(caty-1))
for (k in 2:cat2[j])
ie1[,,catm[j],l,q1]=ie1[,,catm[j],l]+(mu_Mc1[,,k-1]/sum_Mc1-mu_Mc0[,,k-1]/sum_Mc0)*med0$BUGSoutput$sims.list$beta[,q1,catm1[j,1]+k-2]
for (q1 in 1:(caty-1))
aie1[,catm[j],l,q1]<-apply(ie1[,,catm[j],l,q1],1,mean)
}}
if(l%in%data0$contpred1)
{tt1=match(l,data0$contpred1)
for (q1 in 1:(caty-1))
de1[,l,q1]=as.matrix(med0$BUGSoutput$sims.list$c[,q1,data0$contpred3[tt1,1]:data0$contpred3[tt1,2]])%*%
apply(as.matrix(data0$pred.cont.der[,data0$contpred3[tt1,1]:data0$contpred3[tt1,2]]),2,mean)
tt2=tt2+data0$contpred3[tt1,2]-data0$contpred3[tt1,1]+1
}
else
{for (q1 in 1:(caty-1))
de1[,l,q1]=med0$BUGSoutput$sims.list$c[,q1,tt2]
tt2=tt2+1}
#method2: the same for binary and continuous predictors
if(p1>0){
if(p1==1 & c1==1 & contm1[1,1]==contm1[1,2])
{temp.M3=M1
temp.M3[,contm1[1,1]]=M3[,contm3[1,1]]
temp.mu1<-array(0,c(N1,N,caty))
temp.mu3<-temp.mu1
for(q1 in 2:caty)
{temp.mu1[,,q1]=med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x)+med0$BUGSoutput$sims.list$beta[,q1-1,]%*%t(M1)+matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,nrow(x))
temp.mu3[,,q1]=med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x)+med0$BUGSoutput$sims.list$beta[,q1-1,]%*%t(temp.M3)+matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,nrow(x))
if(!is.null(cova))
{temp.mu1[,,q1]=temp.mu1[,,q1]+med0$BUGSoutput$sims.list$eta[,q1-1,]%*%t(cova)
temp.mu3[,,q1]=temp.mu3[,,q1]+med0$BUGSoutput$sims.list$eta[,q1-1,]%*%t(cova)}}
temp.mu1=exp(temp.mu1)
temp.mu3=exp(temp.mu3)
temp.mu1.sum=apply(temp.mu1,c(1,2),sum)
temp.mu3.sum=apply(temp.mu3,c(1,2),sum)
for(q1 in 1:(caty-1))
{ie2[,,contm[1],l,q1]=(med0$BUGSoutput$sims.list$alpha1.a[,1]/deltam[contm[1]])*(temp.mu3[,,q1+1]/temp.mu3.sum-temp.mu1[,,q1+1]/temp.mu1.sum)
aie2[,contm[1],l,q1]=apply(ie2[,,contm[1],l,q1],1,mean,na.rm=T)}
}
else
for (j in 1:p1){
temp.M3=M1
temp.M3[,contm1[j,1]:contm1[j,2]]=M3[,contm3[j,1]:contm3[j,2]]
temp.mu1<-array(0,c(N1,N,caty))
temp.mu3<-temp.mu1
for(q1 in 2:caty)
{temp.mu1[,,q1]=med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x)+med0$BUGSoutput$sims.list$beta[,q1-1,]%*%t(M1)+matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,nrow(x))
temp.mu3[,,q1]=med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x)+med0$BUGSoutput$sims.list$beta[,q1-1,]%*%t(temp.M3)+matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,nrow(x))
if(!is.null(cova))
{temp.mu1[,,q1]=temp.mu1[,,q1]+med0$BUGSoutput$sims.list$eta[,q1-1,]%*%t(cova)
temp.mu3[,,q1]=temp.mu3[,,q1]+med0$BUGSoutput$sims.list$eta[,q1-1,]%*%t(cova)}}
temp.mu1=exp(temp.mu1)
temp.mu3=exp(temp.mu3)
temp.mu1.sum=apply(temp.mu1,c(1,2),sum)
temp.mu3.sum=apply(temp.mu3,c(1,2),sum)
for(q1 in 1:(caty-1))
{ie2[,,contm[j],l,q1]=(med0$BUGSoutput$sims.list$alpha1.a[,j,l]/deltam[contm[j]])*(temp.mu3[,,q1+1]/temp.mu3.sum-temp.mu1[,,q1+1]/temp.mu1.sum)
aie2[,contm[j],l,q1]=apply(ie2[,,contm[j],l,q1],1,mean,na.rm=T)}
}}
#for binary and categorical mediators, method 2 and method 1 are not the same
if(p2>0){
if(p2==1 & c1==1){
temp.M3=M1
temp.M1=M1
temp.M3[,binm1[1]]=1
temp.M1[,binm1[1]]=0
temp.mu1<-array(0,c(N1,N,caty))
temp.mu3<-temp.mu1
for(q1 in 2:caty)
{temp.mu1[,,q1]=med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x)+med0$BUGSoutput$sims.list$beta[,q1-1,]%*%t(temp.M1)+matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,nrow(x))
temp.mu3[,,q1]=med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x)+med0$BUGSoutput$sims.list$beta[,q1-1,]%*%t(temp.M3)+matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,nrow(x))
if(!is.null(cova))
{temp.mu1[,,q1]=temp.mu1[,,q1]+med0$BUGSoutput$sims.list$eta[,q1-1,]%*%t(cova)
temp.mu3[,,q1]=temp.mu3[,,q1]+med0$BUGSoutput$sims.list$eta[,q1-1,]%*%t(cova)}}
temp.mu1=exp(temp.mu1)
temp.mu3=exp(temp.mu3)
temp.mu1.sum=apply(temp.mu1,c(1,2),sum)
temp.mu3.sum=apply(temp.mu3,c(1,2),sum)
bpart=array(0,c(N1,N,caty))
for(q1 in 1:(caty-1))
bpart[,,q1]=temp.mu3[,,q1+1]/temp.mu3.sum-temp.mu1[,,q1+1]/temp.mu1.sum
if (nmc==1){
temp.x1=med0$BUGSoutput$sims.list$alpha0.b[,1]+matrix(med0$BUGSoutput$sims.list$alpha1.b[,1])%*%t(x1.temp)
temp.x3=med0$BUGSoutput$sims.list$alpha0.b[,1]+matrix(med0$BUGSoutput$sims.list$alpha1.b[,1])%*%t(x3.temp)
}
else
{temp.x1=med0$BUGSoutput$sims.list$alpha0.b[,1,mind[binm[1],]]%*%t(mcov[,mind[binm[1],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,1])%*%t(x1.temp)
temp.x3=med0$BUGSoutput$sims.list$alpha0.b[,1,mind[binm[1],]]%*%t(mcov[,mind[binm[1],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,1])%*%t(x3.temp)}
for(q1 in 1:(caty-1)){
ie2[,,binm[1],1,q1]=bpart[,,q1]*(exp(temp.x3)/(1+exp(temp.x3))-exp(temp.x1)/(1+exp(temp.x1)))/deltap[l]
aie2[,binm[1],1,q1] <-apply(ie2[,,binm[1],1,q1],1,mean,na.rm=T)}}
else
for (k in 1:p2){
temp.M3=M1
temp.M1=M1
temp.M3[,binm1[k]]=1
temp.M1[,binm1[k]]=0
temp.mu1<-array(0,c(N1,N,caty))
temp.mu3<-temp.mu1
for(q1 in 2:caty)
{temp.mu1[,,q1]=med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x)+med0$BUGSoutput$sims.list$beta[,q1-1,]%*%t(temp.M1)+matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,nrow(x))
temp.mu3[,,q1]=med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x)+med0$BUGSoutput$sims.list$beta[,q1-1,]%*%t(temp.M3)+matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,nrow(x))
if(!is.null(cova))
{temp.mu1[,,q1]=temp.mu1[,,q1]+med0$BUGSoutput$sims.list$eta[,q1-1,]%*%t(cova)
temp.mu3[,,q1]=temp.mu3[,,q1]+med0$BUGSoutput$sims.list$eta[,q1-1,]%*%t(cova)}}
temp.mu1=exp(temp.mu1)
temp.mu3=exp(temp.mu3)
temp.mu1.sum=apply(temp.mu1,c(1,2),sum)
temp.mu3.sum=apply(temp.mu3,c(1,2),sum)
bpart=array(0,c(N1,N,caty-1))
for(q1 in 1:(caty-1))
bpart[,,q1]=temp.mu3[,,q1+1]/temp.mu3.sum-temp.mu1[,,q1+1]/temp.mu1.sum
if(nmc==1 & p2==1){
temp.x1=med0$BUGSoutput$sims.list$alpha0.b[,k]+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,k,])%*%t(x1.temp)
temp.x3=med0$BUGSoutput$sims.list$alpha0.b[,k]+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,k,])%*%t(x3.temp)
}
else{
temp.x1=med0$BUGSoutput$sims.list$alpha0.b[,k,mind[binm[k],]]%*%t(mcov[,mind[binm[k],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,k,])%*%t(x1.temp)
temp.x3=med0$BUGSoutput$sims.list$alpha0.b[,k,mind[binm[k],]]%*%t(mcov[,mind[binm[k],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,k,])%*%t(x3.temp)
}
for(q1 in 1:(caty-1)){
ie2[,,binm[k],l,q1]=bpart[,,q1]*(exp(temp.x3)/(1+exp(temp.x3))-exp(temp.x1)/(1+exp(temp.x1)))/deltap[l]
aie2[,binm[k],l,q1] <- apply(ie2[,,binm[k],l,q1],1,mean,na.rm=T)}
}
}
if(p3>0){
for (j in 1:p3){
bpart=array(0,c(N1,N,cat2[j]-1,caty-1))
M1.temp=M1
M1.temp[,catm1[j,1]:catm1[j,2]]=0
for (k in catm1[j,1]:catm1[j,2]){
temp.M3=M1.temp
temp.M1=M1.temp
temp.M3[,k]=1
temp.mu1<-array(0,c(N1,N,caty))
temp.mu3<-temp.mu1
for(q1 in 2:caty)
{temp.mu1[,,q1]=med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x)+med0$BUGSoutput$sims.list$beta[,q1-1,]%*%t(temp.M1)+matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,nrow(x))
temp.mu3[,,q1]=med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x)+med0$BUGSoutput$sims.list$beta[,q1-1,]%*%t(temp.M3)+matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,nrow(x))
if(!is.null(cova))
{temp.mu1[,,q1]=temp.mu1[,,q1]+med0$BUGSoutput$sims.list$eta[,q1-1,]%*%t(cova)
temp.mu3[,,q1]=temp.mu3[,,q1]+med0$BUGSoutput$sims.list$eta[,q1-1,]%*%t(cova)}}
temp.mu1=exp(temp.mu1)
temp.mu3=exp(temp.mu3)
temp.mu1.sum=apply(temp.mu1,c(1,2),sum)
temp.mu3.sum=apply(temp.mu3,c(1,2),sum)
for(q1 in 1:(caty-1))
bpart[,,k-catm1[j,1]+1,q1]=exp(temp.mu3)/(1+exp(temp.mu3))-exp(temp.mu1)/(1+exp(temp.mu1))
}
mu_Mc1<-array(0,c(N1,N,cat2[j]-1))
mu_Mc0<-array(0,c(N1,N,cat2[j]-1))
for (k in 2:cat2[j]){
mu_Mc1[,,k-1] <- exp(as.matrix(med0$BUGSoutput$sims.list$alpha0.c[,j,k-1,mind[catm[j],]])%*%t(mcov[,mind[catm[j],]])+med0$BUGSoutput$sims.list$alpha1.c[,j,k-1,]%*%t(x3.temp))
mu_Mc0[,,k-1] <- exp(as.matrix(med0$BUGSoutput$sims.list$alpha0.c[,j,k-1,mind[catm[j],]])%*%t(mcov[,mind[catm[j],]])+med0$BUGSoutput$sims.list$alpha1.c[,j,k-1,]%*%t(x1.temp))
}
sum_Mc1 <-apply(mu_Mc1,c(1,2),sum)+1
sum_Mc0 <-apply(mu_Mc0,c(1,2),sum)+1
for(q1 in 1:(caty-1))
{for (k in 2:cat2[j])
ie2[,,catm[j],l,q1]=ie2[,,catm[j],l,q1]+(mu_Mc1[,,k-1]/sum_Mc1-mu_Mc0[,,k-1]/sum_Mc0)*bpart[,,k-1,q1]
aie2[,catm[j],l,q1]<-apply(ie2[,,catm[j],l,q1],1,mean,na.rm=T)}
}
}
temp.x1=x
temp.x3=x
if(l%in%data0$contpred1.0)
{tt1=match(l,data0$contpred1.0)
temp.x3[,data0$contpred2[tt1,1]:data0$contpred2[tt1,2]]=data0$pred3[,data0$contpred3[tt1,1]:data0$contpred3[tt1,2]]}
else if(l%in%data0$binpred1.0)
{tt1=match(l,data0$binpred1)
temp.x1[,data0$binpred2[tt1]]=0
temp.x3[,data0$binpred2[tt1]]=1
deltap[l]=1}
else
{for (i in 1:nrow(data0$catpred1.0))
if(l%in%(data0$catpred1.0[i,1]:data0$catpred1.0[i,2]))
tt1=i
temp.x1[,data0$catpred2[tt1,1]:data0$catpred2[tt1,2]]=0
temp.x3[,data0$catpred2[tt1,1]:data0$catpred2[tt1,2]]=0
temp.x3[,l]=1
deltap[l]=1}
temp.mu1<-array(0,c(N1,N,caty))
temp.mu3<-temp.mu1
for(q1 in 2:caty)
{temp.mu1[,,q1]=med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(temp.x1)+med0$BUGSoutput$sims.list$beta[,q1-1,]%*%t(M1)+matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,nrow(x))
temp.mu3[,,q1]=med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(temp.x3)+med0$BUGSoutput$sims.list$beta[,q1-1,]%*%t(M1)+matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,nrow(x))
if(!is.null(cova))
{temp.mu1[,,q1]=temp.mu1[,,q1]+med0$BUGSoutput$sims.list$eta[,q1-1,]%*%t(cova)
temp.mu3[,,q1]=temp.mu3[,,q1]+med0$BUGSoutput$sims.list$eta[,q1-1,]%*%t(cova)}}
temp.mu1=exp(temp.mu1)
temp.mu3=exp(temp.mu3)
temp.mu1.sum=apply(temp.mu1,c(1,2),sum)
temp.mu3.sum=apply(temp.mu3,c(1,2),sum)
for(q1 in 1:(caty-1))
{de2.1=(temp.mu3[,,q1+1]/temp.mu3.sum-temp.mu1[,,q1+1]/temp.mu1.sum)/deltap[l]
de2[,l,q1]=apply(de2.1,1,mean,na.rm=T)}
#method 3:parametric
#3.1. get M1(x) and M1(x+dx) for y
if(p1>0){
if(c1==1 & p1+p2+p3==1 & contm1[1,1]==contm1[1,2]){
if(nmc==1)
{mu_M2[,contm1[1,1],] <- med0$BUGSoutput$sims.list$alpha0[,contm1[1,1]]+as.matrix(med0$BUGSoutput$sims.list$alpha1[,contm1[1,1]])%*%t(x1.temp)
mu_M3[,contm1[1,1],] <- med0$BUGSoutput$sims.list$alpha0[,contm1[1,1]]+as.matrix(med0$BUGSoutput$sims.list$alpha1[,contm1[1,1]])%*%t(x3.temp)}
else
{mu_M2[,contm1[1,1],] <- med0$BUGSoutput$sims.list$alpha0[,contm1[1,1],mind[contm[1],]]%*%t(mcov[,mind[contm[1],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1[,contm1[1,1]])%*%t(x1.temp)
mu_M3[,contm1[1,1],] <- med0$BUGSoutput$sims.list$alpha0[,contm1[1,1],mind[contm[1],]]%*%t(mcov[,mind[contm[1],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1[,contm1[1,1]])%*%t(x3.temp)}
}
else
for (j in 1:p1){
for (k in contm1[j,1]:contm1[j,2]){
if(nmc==1 & p1+p2+p3==1)
{mu_M2[,k,] <- med0$BUGSoutput$sims.list$alpha0[,k]+med0$BUGSoutput$sims.list$alpha1[,k,l]%*%t(x1.temp)
mu_M3[,k,] <- med0$BUGSoutput$sims.list$alpha0[,k]+med0$BUGSoutput$sims.list$alpha1[,k,l]%*%t(x3.temp)}
else
{mu_M2[,k,] <- med0$BUGSoutput$sims.list$alpha0[,k,mind[contm[j],]]%*%t(mcov[,mind[contm[j],]])+med0$BUGSoutput$sims.list$alpha1[,k,]%*%t(x1.temp)
mu_M3[,k,] <- med0$BUGSoutput$sims.list$alpha0[,k,mind[contm[j],]]%*%t(mcov[,mind[contm[j],]])+med0$BUGSoutput$sims.list$alpha1[,k,]%*%t(x3.temp)}
}}}
if(p2>0){
if(p2==1 & c1==1)
{if(nmc==1)
{temp.x1=med0$BUGSoutput$sims.list$alpha0.b[,1]+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,1])%*%t(x1.temp)
temp.x3=med0$BUGSoutput$sims.list$alpha0.b[,1]+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,1])%*%t(x3.temp)}
else
{temp.x1=med0$BUGSoutput$sims.list$alpha0.b[,1,mind[binm[1],]]%*%t(mcov[,mind[binm[1],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,1])%*%t(x1.temp)
temp.x3=med0$BUGSoutput$sims.list$alpha0.b[,1,mind[binm[1],]]%*%t(mcov[,mind[binm[1],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,1])%*%t(x3.temp)}
mu_M2[,binm[1],] <- exp(temp.x1)/(1+exp(temp.x1))
mu_M3[,binm[1],] <- exp(temp.x3)/(1+exp(temp.x3))
}
else{
for (k in 1:p2){
if(nmc==1 & p2==1)
{temp.x1=med0$BUGSoutput$sims.list$alpha0.b[,k]+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,k,])%*%t(x1.temp)
temp.x3=med0$BUGSoutput$sims.list$alpha0.b[,k]+med0$BUGSoutput$sims.list$alpha1.b[,k,]%*%t(x3.temp)}
else
{temp.x1=med0$BUGSoutput$sims.list$alpha0.b[,k,mind[binm[k],]]%*%t(mcov[,mind[binm[k],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,k,])%*%t(x1.temp)
temp.x3=med0$BUGSoutput$sims.list$alpha0.b[,k,mind[binm[k],]]%*%t(mcov[,mind[binm[k],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,k,])%*%t(x3.temp)}
mu_M2[,binm1[k],] <- exp(temp.x1)/(1+exp(temp.x1))
mu_M3[,binm1[k],] <- exp(temp.x3)/(1+exp(temp.x3))}
}}
if(p3>0){
for (j in 1:p3){
mu_Mc1<-array(0,c(N1,N,cat2[j]-1))
mu_Mc0<-array(0,c(N1,N,cat2[j]-1))
for (k in 2:cat2[j]){
mu_Mc1[,,k-1] <- exp(as.matrix(med0$BUGSoutput$sims.list$alpha0.c[,j,k-1,mind[catm[j],]])%*%t(mcov[,mind[catm[j],]])+med0$BUGSoutput$sims.list$alpha1.c[,j,k-1,]%*%t(x3.temp))
mu_Mc0[,,k-1] <- exp(as.matrix(med0$BUGSoutput$sims.list$alpha0.c[,j,k-1,mind[catm[j],]])%*%t(mcov[,mind[catm[j],]])+med0$BUGSoutput$sims.list$alpha1.c[,j,k-1,]%*%t(x1.temp))
}
sum_Mc1 <-apply(mu_Mc1,c(1,2),sum)+1
sum_Mc0 <-apply(mu_Mc0,c(1,2),sum)+1
for (k in 2:cat2[j])
{mu_M2[,catm1[j,1]+k-2,]=mu_Mc0[,,k-1]/sum_Mc0
mu_M3[,catm1[j,1]+k-2,]=mu_Mc1[,,k-1]/sum_Mc1}
}}
#3.2. get x and dx for y
x1.temp1=x
x3.temp1=x
if(l%in%as.vector(data0$contpred1.0)){ #need the continuous predictor in its original format
i=match(l,data0$contpred1.0)
x3.temp1[,data0$contpred2[i,1]:data0$contpred2[i,2]]=data0$pred3[,data0$contpred3[i,1]:data0$contpred3[i,2]]
}
else if(l%in%data0$binpred1.0)
{i=match(l,data0$binpred1.0)
x1.temp1[,data0$binpred2[i]]=0
x3.temp1[,data0$binpred2[i]]=1
deltap[l]=1}
else{ #categorical predictor
for (i in 1:nrow(data0$catpred1.0))
if(l%in%(data0$catpred1.0[i,1]:data0$catpred1.0[i,2]))
{x1.temp1[,data0$catpred2[i,1]:data0$catpred2[i,2]]=0
x3.temp1[,data0$catpred2[i,1]:data0$catpred2[i,2]]=0
di=match(l,data0$catpred1.0[i,1]:data0$catpred1.0[i,2])
x3.temp1[,data0$catpred2[i,1]+di-1]=1}
deltap[l]=1
}
#3.3. get the total effect
mu_y0<-array(0,c(N1,N,caty))
mu_y1<-mu_y0
for(q1 in 2:caty)
{temp1=array(0,c(nrow(M1)+1,ncol(M1),N1))
temp1[2:(nrow(M1)+1),,]=mu_M2
temp1[1,,]=t(med0$BUGSoutput$sims.list$beta[,q1-1,])
temp2=temp1
temp2[2:(nrow(M1)+1),,]=mu_M3
if(is.null(cova))
{mu_y0[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x1.temp1) + t(apply(temp1,3,matrix.prod))
mu_y1[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x3.temp1) + t(apply(temp2,3,matrix.prod))}
else
{mu_y0[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x1.temp1) + med0$BUGSoutput$sims.list$eta[,q1-1,]%*%t(as.matrix(cova))+t(apply(temp1,3,matrix.prod))
mu_y1[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x3.temp1) + med0$BUGSoutput$sims.list$eta[,q1-1,]%*%t(as.matrix(cova))+t(apply(temp2,3,matrix.prod))
}} #get the linear part
mu_y0=exp(mu_y0)
mu_y1=exp(mu_y1)
mu_y0.sum=apply(mu_y0,c(1,2),sum)
mu_y1.sum=apply(mu_y1,c(1,2),sum)
for(q1 in 1:(caty-1)){
te3[,,q1]=(mu_y1[,,q1+1]/mu_y1.sum-mu_y0[,,q1+1]/mu_y0.sum)/deltap[l]
ate3[,l,q1]=apply(te3[,,q1],1,mean,na.rm=T)}
#3.4. calculate the ie
j1=sample(1:N,size=N*N1,replace=T)
j2=sample(1:N,size=N*N1,replace=T)
#3.4.1. continuous mediators
if(p1>0){
for (j in 1:p1){
mu_y0.2<-array(0,c(N1,N,caty))
mu_y1.2<-mu_y0.2
for(q1 in 2:caty)
{temp1.1=temp1
temp1.2=temp2
temp1.1[1,,]=t(med0$BUGSoutput$sims.list$beta[,q1-1,])
temp1.2[1,,]=t(med0$BUGSoutput$sims.list$beta[,q1-1,])
for (i in contm1[j,1]:contm1[j,2])
{temp1.1[2:(nrow(M1)+1),i,]=matrix(M1[j1,i],N,N1)
temp1.2[2:(nrow(M1)+1),i,]=matrix(M1[j1,i],N,N1)} #j2/j1
if(is.null(cova))
{mu_y0.2[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x1.temp1) +t(apply(temp1.1,3,matrix.prod))
mu_y1.2[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x3.temp1) +t(apply(temp1.2,3,matrix.prod)) }
else{
mu_y0.2[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x1.temp1) + med0$BUGSoutput$sims.list$eta[,q1-1,]%*%t(cova)+t(apply(temp1.1,3,matrix.prod))
mu_y1.2[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x3.temp1) + med0$BUGSoutput$sims.list$eta[,q1-1,]%*%t(cova)+t(apply(temp1.2,3,matrix.prod))}
}
mu_y0.2=exp(mu_y0.2)
mu_y1.2=exp(mu_y1.2)
mu_y0.2.sum=apply(mu_y0.2,c(1,2),sum)
mu_y1.2.sum=apply(mu_y1.2,c(1,2),sum)
for(q1 in 1:(caty-1))
ie3[,,contm[j],l,q1]<-te3[,,q1]-(mu_y1.2[,,q1+1]/mu_y1.2.sum-mu_y0.2[,,q1+1]/mu_y0.2.sum)/deltap[l]
}}
#3.4.2. binary mediators
if(p2>0){
for (k in 1:p2){
mu_y0.2<-array(0,c(N1,N,caty))
mu_y1.2<-mu_y0.2
for(q1 in 2:caty)
{temp1.1=temp1
temp1.2=temp2
temp1.1[1,,]=t(med0$BUGSoutput$sims.list$beta[,q1-1,])
temp1.2[1,,]=t(med0$BUGSoutput$sims.list$beta[,q1-1,])
temp1.1[2:(nrow(M1)+1),binm1[k],]=matrix(M1[j1,binm1[k]],N,N1)
temp1.2[2:(nrow(M1)+1),binm1[k],]=matrix(M1[j1,binm1[k]],N,N1) #j2/j1
if (is.null(cova)){
mu_y0.2[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x1.temp1) + t(apply(temp1.1,3,matrix.prod))
mu_y1.2[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x3.temp1) + t(apply(temp1.2,3,matrix.prod))}
else{
mu_y0.2[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x1.temp1) + med0$BUGSoutput$sims.list$eta[,q1-1,]%*%t(cova)+t(apply(temp1.1,3,matrix.prod))
mu_y1.2[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x3.temp1) + med0$BUGSoutput$sims.list$eta[,q1-1,]%*%t(cova)+t(apply(temp1.2,3,matrix.prod))}
}
mu_y0.2=exp(mu_y0.2)
mu_y1.2=exp(mu_y1.2)
mu_y0.2.sum=apply(mu_y0.2,c(1,2),sum)
mu_y1.2.sum=apply(mu_y1.2,c(1,2),sum)
for(q1 in 1:(caty-1))
ie3[,,binm[k],l,q1]<-te3[,,q1]-(mu_y1.2[,,q1+1]/mu_y1.2.sum-mu_y0.2[,,q1+1]/mu_y0.2.sum)/deltap[l]
}}
if(p3>0){
for (j in 1:p3){
mu_y0.2<-array(0,c(N1,N,caty))
mu_y1.2<-mu_y0.2
for(q1 in 2:caty)
{temp1.1=temp1
temp1.2=temp2
temp1.1[1,,]=t(med0$BUGSoutput$sims.list$beta[,q1-1,])
temp1.2[1,,]=t(med0$BUGSoutput$sims.list$beta[,q1-1,])
for (i in catm1[j,1]:catm1[j,2])
{temp1.1[2:(nrow(M1)+1),i,]=matrix(M1[j1,i],N,N1)
temp1.2[2:(nrow(M1)+1),i,]=matrix(M1[j1,i],N,N1)} #j2/j1
if(is.null(cova)){
mu_y0.2[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x1.temp1) + t(apply(temp1.1,3,matrix.prod))
mu_y1.2[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x3.temp1) + t(apply(temp1.2,3,matrix.prod))}
else{
mu_y0.2[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x1.temp1) + med0$BUGSoutput$sims.list$eta[,q1-1,]%*%t(cova)+t(apply(temp1.1,3,matrix.prod))
mu_y1.2[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x3.temp1) + med0$BUGSoutput$sims.list$eta[,q1-1,]%*%t(cova)+t(apply(temp1.2,3,matrix.prod))}
}
mu_y0.2=exp(mu_y0.2)
mu_y1.2=exp(mu_y1.2)
mu_y0.2.sum=apply(mu_y0.2,c(1,2),sum)
mu_y1.2.sum=apply(mu_y1.2,c(1,2),sum)
for(q1 in 1:(caty-1))
ie3[,,catm[j],l,q1]<-te3[,,q1]-(mu_y1.2[,,q1+1]/mu_y1.2.sum-mu_y0.2[,,q1+1]/mu_y0.2.sum)/deltap[l]
}}
aie3[,,l,]<-apply(array(ie3[,,,l,],c(N1,N,p1+p2+p3,caty-1)),c(1,3,4),mean,na.rm=T)
#3.5. Calculate the de
mu_y0.2<-array(0,c(N1,N,caty))
mu_y1.2<-mu_y0.2
for(q1 in 2:caty)
{temp1.1=temp1
temp1.2=temp2
temp1.1[1,,]=t(med0$BUGSoutput$sims.list$beta[,q1-1,])
temp1.2[1,,]=t(med0$BUGSoutput$sims.list$beta[,q1-1,])
for (i in 1:ncol(M1))
{temp1.1[2:(nrow(M1)+1),i,]=matrix(M1[j1,i],N,N1)
temp1.2[2:(nrow(M1)+1),i,]=matrix(M1[j1,i],N,N1)} #j2/j1
if(is.null(cova)){
mu_y0.2[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x1.temp1) + t(apply(temp1.1,3,matrix.prod))
mu_y1.2[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x3.temp1) + t(apply(temp1.2,3,matrix.prod))}
else{
mu_y0.2[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x1.temp1) + med0$BUGSoutput$sims.list$eta[,q1-1,]%*%t(cova)+t(apply(temp1.1,3,matrix.prod))
mu_y1.2[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x3.temp1) + med0$BUGSoutput$sims.list$eta[,q1-1,]%*%t(cova)+t(apply(temp1.2,3,matrix.prod))}
}
mu_y0.2=exp(mu_y0.2)
mu_y1.2=exp(mu_y1.2)
mu_y0.2.sum=apply(mu_y0.2,c(1,2),sum)
mu_y1.2.sum=apply(mu_y1.2,c(1,2),sum)
for(q1 in 1:(caty-1))
{de3[,,q1]<-(mu_y1.2[,,q1+1]/mu_y1.2.sum-mu_y0.2[,,q1+1]/mu_y0.2.sum)/deltap[l]
ade3[,l,q1]=apply(de3[,,q1],1,mean,na.rm=T)}
#method3: semi-parametric for binary or categorical predictors
if(!l%in%data0$contpred1.0){
if(!is.null(data0$binpred1.0))
{if (l%*%data0$binpred1.0)
{M.0=data.frame(M1[x1[,l]==0,])
y.0=y[x1[,l]==0]}
else
{for(i in 1:nrow(data0$catpred1.0))
if(l%in%(data0$catpred1.0[i,1]:data0$catpred1.0[i,2]))
tt1=i
M.0=data.frame(M1[apply(x1[,data0$catpred1.0[tt1,1]:data0$catpred1.0[[tt1,2]]]==1,1,sum)==0,])
y.0=y[apply(x1[,data0$catpred1.0[tt1,1]:data0$catpred1.0[[tt1,2]]]==1,1,sum)==0]}}
else
{for(i in 1:nrow(data0$catpred1.0))
if(l%in%(data0$catpred1.0[i,1]:data0$catpred1.0[i,2]))
tt1=i
M.0=data.frame(M1[apply(x1[,data0$catpred1.0[tt1,1]:data0$catpred1.0[[tt1,2]]]==1,1,sum)==0,])
y.0=y[apply(x1[,data0$catpred1.0[tt1,1]:data0$catpred1.0[[tt1,2]]]==1,1,sum)==0]}
y.1=y[x1[,l]==1]
M.1=data.frame(M1[x1[,l]==1,])
j1=sample(1:N,size=N*N1,replace=T)
j2=sample(1:N,size=N*N1,replace=T)
n3=nrow(M.0)
n4=nrow(M.1)
j3=sample(1:n3,size=N*N1,replace = T)
j4=sample(1:n4,size=N*N1,replace = T)
for (i in 1:ncol(M1))
{mu.M0[,i,]=matrix(M.0[j3,i],N,N1)
mu.M1[,i,]=matrix(M.1[j4,i],N,N1)}
#4.1. get the total effect
mu_y0=matrix(y.0[j3],N1,N)
mu_y1=matrix(y.1[j4],N1,N)
temp1=array(0,c(nrow(M1)+1,ncol(M1),N1))
temp1[2:(nrow(M1)+1),,]=mu.M0
temp2=temp1
temp2[2:(nrow(M1)+1),,]=mu.M1
temp.levels=levels(y)[-1]
for(q1 in 1:length(temp.levels))
{te4[,,q1]<- (mu_y1==temp.levels[q1])-(mu_y0==temp.levels[q1])
ate4[,l,q1]=apply(te4[,,q1],1,mean,na.rm=T)}
#4.2. Get the ies
#ie for continuous mediators
if(p1>0){
for (j in 1:p1){
mu_y0.2<-array(0,c(N1,N,caty))
mu_y1.2<-mu_y0.2
for(q1 in 2:caty)
{temp1.1=temp1
temp1.2=temp2
temp1.1[1,,]=t(med0$BUGSoutput$sims.list$beta[,q1-1,])
temp1.2[1,,]=t(med0$BUGSoutput$sims.list$beta[,q1-1,])
for (i in contm1[j,1]:contm1[j,2])
{temp1.1[2:(nrow(M1)+1),i,]=matrix(M1[j1,i],N,N1)
temp1.2[2:(nrow(M1)+1),i,]=matrix(M1[j1,i],N,N1)} #j2/j1
if(is.null(cova))
{mu_y0.2[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x1.temp1) +t(apply(temp1.1,3,matrix.prod))
mu_y1.2[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x3.temp1) +t(apply(temp1.2,3,matrix.prod)) }
else{
mu_y0.2[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x1.temp1) + med0$BUGSoutput$sims.list$eta[,q1-1,]%*%t(cova)+t(apply(temp1.1,3,matrix.prod))
mu_y1.2[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x3.temp1) + med0$BUGSoutput$sims.list$eta[,q1-1,]%*%t(cova)+t(apply(temp1.2,3,matrix.prod))}
}
mu_y0.2=exp(mu_y0.2)
mu_y1.2=exp(mu_y1.2)
mu_y0.2.sum=apply(mu_y0.2,c(1,2),sum)
mu_y1.2.sum=apply(mu_y1.2,c(1,2),sum)
for(q1 in 1:(caty-1))
ie4[,,contm[j],l,q1]<-te4[,,q1]-(mu_y1.2[,,q1+1]/mu_y1.2.sum-mu_y0.2[,,q1+1]/mu_y0.2.sum)/deltap[l]
}}
#ie for binary mediators
if(p2>0){
for (k in 1:p2){
mu_y0.2<-array(0,c(N1,N,caty))
mu_y1.2<-mu_y0.2
for(q1 in 2:caty)
{temp1.1=temp1
temp1.2=temp2
temp1.1[2:(nrow(M1)+1),binm1[k],]=matrix(M1[j1,binm1[k]],N,N1)
temp1.2[2:(nrow(M1)+1),binm1[k],]=matrix(M1[j1,binm1[k]],N,N1) #j2/j1
temp1.1[1,,]=t(med0$BUGSoutput$sims.list$beta[,q1-1,])
temp1.2[1,,]=t(med0$BUGSoutput$sims.list$beta[,q1-1,])
if (is.null(cova)){
mu_y0.2[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x1.temp1) + t(apply(temp1.1,3,matrix.prod))
mu_y1.2[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x3.temp1) + t(apply(temp1.2,3,matrix.prod))}
else{
mu_y0.2[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x1.temp1) + med0$BUGSoutput$sims.list$eta[,q1-1,]%*%t(cova)+t(apply(temp1.1,3,matrix.prod))
mu_y1.2[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x3.temp1) + med0$BUGSoutput$sims.list$eta[,q1-1,]%*%t(cova)+t(apply(temp1.2,3,matrix.prod))}
}
mu_y0.2=exp(mu_y0.2)
mu_y1.2=exp(mu_y1.2)
mu_y0.2.sum=apply(mu_y0.2,c(1,2),sum)
mu_y1.2.sum=apply(mu_y1.2,c(1,2),sum)
for(q1 in 1:(caty-1))
ie4[,,binm[k],l,q1]<-te4[,,q1]-(mu_y1.2[,,q1+1]/mu_y1.2.sum-mu_y0.2[,,q1+1]/mu_y0.2.sum)
}}
#ie for categorical mediators
if(p3>0){
for (j in 1:p3){
mu_y0.2<-array(0,c(N1,N,caty))
mu_y1.2<-mu_y0.2
for(q1 in 2:caty)
{temp1.1=temp1
temp1.2=temp2
temp1.1[1,,]=t(med0$BUGSoutput$sims.list$beta[,q1-1,])
temp1.2[1,,]=t(med0$BUGSoutput$sims.list$beta[,q1-1,])
for (i in catm1[j,1]:catm1[j,2])
{temp1.1[2:(nrow(M1)+1),i,]=matrix(M1[j1,i],N,N1)
temp1.2[2:(nrow(M1)+1),i,]=matrix(M1[j1,i],N,N1)} #j2/j1
if(is.null(cova)){
mu_y0.2[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x1.temp1) + t(apply(temp1.1,3,matrix.prod))
mu_y1.2[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x3.temp1) + t(apply(temp1.2,3,matrix.prod))}
else{
mu_y0.2[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x1.temp1) + med0$BUGSoutput$sims.list$eta[,q1-1,]%*%t(cova)+t(apply(temp1.1,3,matrix.prod))
mu_y1.2[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x3.temp1) + med0$BUGSoutput$sims.list$eta[,q1-1,]%*%t(cova)+t(apply(temp1.2,3,matrix.prod))}
}
mu_y0.2=exp(mu_y0.2)
mu_y1.2=exp(mu_y1.2)
mu_y0.2.sum=apply(mu_y0.2,c(1,2),sum)
mu_y1.2.sum=apply(mu_y1.2,c(1,2),sum)
for(q1 in 1:(caty-1))
ie4[,,catm[j],l,q1]<-te4[,,q1]-(mu_y1.2[,,q1+1]/mu_y1.2.sum-mu_y0.2[,,q1+1]/mu_y0.2.sum)
}}
aie4[,,l,]<-apply(array(ie4[,,,l,],c(N1,N,p1+p2+p3,caty-1)),c(1,3,4),mean,na.rm=T)
#4.3. Calculate the de
mu_y0.2<-array(0,c(N1,N,caty))
mu_y1.2<-mu_y0.2
for(q1 in 2:caty)
{temp1.1=temp1
temp1.2=temp2
for (i in 1:ncol(M1))
{temp1.1[2:(nrow(M1)+1),i,]=matrix(M1[j1,i],N,N1)
temp1.2[2:(nrow(M1)+1),i,]=matrix(M1[j1,i],N,N1)} #j2/j1
temp1.1[1,,]=t(med0$BUGSoutput$sims.list$beta[,q1-1,])
temp1.2[1,,]=t(med0$BUGSoutput$sims.list$beta[,q1-1,])
if(is.null(cova)){
mu_y0.2[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x1.temp1) + t(apply(temp1.1,3,matrix.prod))
mu_y1.2[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x3.temp1) + t(apply(temp1.2,3,matrix.prod))}
else{
mu_y0.2[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x1.temp1) + med0$BUGSoutput$sims.list$eta[,q1-1,]%*%t(cova)+t(apply(temp1.1,3,matrix.prod))
mu_y1.2[,,q1]<- matrix(med0$BUGSoutput$sims.list$beta0[,q1-1],N1,N) + med0$BUGSoutput$sims.list$c[,q1-1,]%*%t(x3.temp1) + med0$BUGSoutput$sims.list$eta[,q1-1,]%*%t(cova)+t(apply(temp1.2,3,matrix.prod))}
}
mu_y0.2=exp(mu_y0.2)
mu_y1.2=exp(mu_y1.2)
mu_y0.2.sum=apply(mu_y0.2,c(1,2),sum)
mu_y1.2.sum=apply(mu_y1.2,c(1,2),sum)
}
}
ate1=apply(aie1,c(1,3,4),sum)+de1
ate2=apply(aie2,c(1,3,4),sum)+de2
colnames(aie4)=colnames(M2)
colnames(aie1)=colnames(M2)
colnames(aie2)=colnames(M2)
colnames(aie3)=colnames(M2)
}
else{
aie1=array(0,c(N1,p1+p2+p3,c1))
ie1=array(0,dim=c(N1,N,p1+p2+p3,c1))
tt2=1
de1=NULL
n.cont=1
aie2=array(0,c(N1,p1+p2+p3,c1))
ie2=array(0,dim=c(N1,N,p1+p2+p3,c1))
de2=NULL
ate3=matrix(0,N1,c1)
omu3=ate3
ade3=matrix(0,N1,c1)
aie3=array(0,c(N1,p1+p2+p3,c1))
ie3=array(0,dim=c(N1,N,p1+p2+p3,c1))
mu_M2=array(0,c(dim(M1),N1))
mu_M3=mu_M2
ade4=matrix(0,N1,c1)
ie4=array(0,dim=c(N1,N,p1+p2+p3,c1))
mu.M0=array(0,c(dim(M1),N1))
mu.M1=mu.M0
ate4=matrix(0,N1,c1)
omu4=ate4
aie4=array(0,c(N1,p1+p2+p3,c1))
for (l in 1:c1){
x1.temp=x1
x3.temp=x1
if(l%in%data0$contpred1.0){
x3.temp[,l]=x1[,l]+deltap[l]
}
else if(l%in%data0$binpred1.0)
{x1.temp[,l]=0
x3.temp[,l]=1
deltap[l]=1}
else{ #categorical predictor
for (i in 1:nrow(data0$catpred1.0))
if(l%in%(data0$catpred1.0[i,1]:data0$catpred1.0[i,2]))
{x1.temp[,data0$catpred1.0[i,1]:data0$catpred1.0[i,2]]=0
x3.temp[,data0$catpred1.0[i,1]:data0$catpred1.0[i,2]]=0
x3.temp[,l]=1}
deltap[l]=1
}
#method 1: the same for binary or continuous predictors
if(p1>0){
if (p1+p2+p3==1 & c1==1 & contm1[1,1]==contm1[1,2])
aie1[,contm[1],l]=med0$BUGSoutput$sims.list$alpha1[,1]*med0$BUGSoutput$sims.list$beta[,contm1[1,1]]#*mean(data0$m.cont.der[,contm3[1,1]]) since alpha1 is the coefficient of x on the change of f(M), beta is the coefficient of f(M) on the change of y
else
for (j in 1: p1)
aie1[,contm[j],l]=apply(as.matrix(med0$BUGSoutput$sims.list$alpha1[,contm1[j,1]:contm1[j,2],l])*as.matrix(med0$BUGSoutput$sims.list$beta[,contm1[j,1]:contm1[j,2]]),1,sum)#*mean(data0$m.cont.der[,contm3[j,1]])
}
if(p2>0){
if(p2==1 & c1==1) #since c1=1, the predictor can only be binary or continuous
{if (nmc==1)
{temp.x1=med0$BUGSoutput$sims.list$alpha0.b[,1]+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,1])%*%t(x1.temp)
temp.x3=med0$BUGSoutput$sims.list$alpha0.b[,1]+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,1])%*%t(x3.temp)}
else
{temp.x1=med0$BUGSoutput$sims.list$alpha0.b[,1,mind[binm[1],]]%*%t(mcov[,mind[binm[1],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,1])%*%t(x1.temp)
temp.x3=med0$BUGSoutput$sims.list$alpha0.b[,1,mind[binm[1],]]%*%t(mcov[,mind[binm[1],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,1])%*%t(x3.temp)}
if(!1%in%data0$contpred1.0) #if the predictor is binary
ie1[,,binm[1],1]=med0$BUGSoutput$sims.list$beta[,binm1[1]]*(exp(temp.x3)/(1+exp(temp.x3))-exp(temp.x1)/(1+exp(temp.x1)))
else #if the predictor is continuous
ie1[,,binm[1],1]=med0$BUGSoutput$sims.list$alpha1.b[,1]*med0$BUGSoutput$sims.list$beta[,binm1[1]]*exp(temp.x1)/(1+exp(temp.x1))^2
aie1[,binm[1],l] <-apply(ie1[,,binm[1],1],1,mean)
}
else
for (k in 1:p2){
if (nmc==1 & p2==1)
{temp.x1=med0$BUGSoutput$sims.list$alpha0.b[,k]+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,k,])%*%t(x1.temp)
temp.x3=med0$BUGSoutput$sims.list$alpha0.b[,k]+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,k,])%*%t(x3.temp)
}
else
{temp.x1=med0$BUGSoutput$sims.list$alpha0.b[,k,mind[binm[k],]]%*%t(mcov[,mind[binm[k],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,k,])%*%t(x1.temp)
temp.x3=med0$BUGSoutput$sims.list$alpha0.b[,k,mind[binm[k],]]%*%t(mcov[,mind[binm[k],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,k,])%*%t(x3.temp)
}
if(!l%in%data0$contpred1.0) #if the predictor is binary or categorical
ie1[,,binm[k],l]=med0$BUGSoutput$sims.list$beta[,binm1[k]]*(exp(temp.x3)/(1+exp(temp.x3))-exp(temp.x1)/(1+exp(temp.x1)))
else if(data0$contpred1.0) #for continuous predictor
ie1[,,binm[1],l]=med0$BUGSoutput$sims.list$alpha1.b[,k,l]*med0$BUGSoutput$sims.list$beta[,binm1[k]]*exp(temp.x1)/(1+exp(temp.x1))^2
aie1[,binm[k],l] <- apply(ie1[,,binm[k],l],1,mean)
}
}
if(p3>0){
for (j in 1:p3){
mu_Mc1<-array(0,c(N1,N,cat2[j]-1))
mu_Mc0<-array(0,c(N1,N,cat2[j]-1))
for (k in 2:cat2[j]){
mu_Mc0[,,k-1] <- exp(as.matrix(med0$BUGSoutput$sims.list$alpha0.c[,j,k-1,mind[catm[j],]])%*%t(mcov[,mind[catm[j],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.c[,j,k-1,])%*%t(x1.temp))
mu_Mc1[,,k-1] <- exp(as.matrix(med0$BUGSoutput$sims.list$alpha0.c[,j,k-1,mind[catm[j],]])%*%t(mcov[,mind[catm[j],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.c[,j,k-1,])%*%t(x3.temp))
}
sum_Mc1 <-apply(mu_Mc1,c(1,2),sum)+1
sum_Mc0 <-apply(mu_Mc0,c(1,2),sum)+1
if(l%in%data0$contpred1.0) #for continuous predictor
{tt.0=rep(0,N1)
for (k in 2:cat2[j])
tt.0=mu_Mc0[,,k-1]/sum_Mc0*med0$BUGSoutput$sims.list$alpha1.c[,j,k-1,l]
for (k in 2:cat2[j])
ie1[,,catm[j],l]=ie1[,,catm[j],l]+(mu_Mc0[,,k-1]/sum_Mc0)*med0$BUGSoutput$sims.list$beta[,catm1[j,1]+k-2]*(med0$BUGSoutput$sims.list$alpha1.c[,j,k-1,l]-tt.0)
}
else #for binary or categorical predictor
for (k in 2:cat2[j])
ie1[,,catm[j],l]=ie1[,,catm[j],l]+(mu_Mc1[,,k-1]/sum_Mc1-mu_Mc0[,,k-1]/sum_Mc0)*med0$BUGSoutput$sims.list$beta[,catm1[j,1]+k-2]
aie1[,catm[j],l]<-apply(ie1[,,catm[j],l],1,mean)
}}
if(l%in%data0$contpred1)
{tt1=match(l,data0$contpred1)
de1=cbind(de1,as.matrix(med0$BUGSoutput$sims.list$c[,data0$contpred3[tt1,1]:data0$contpred3[tt1,2]])%*%
apply(as.matrix(data0$pred.cont.der[,data0$contpred3[tt1,1]:data0$contpred3[tt1,2]]),2,mean))
tt2=tt2+data0$contpred3[tt1,2]-data0$contpred3[tt1,1]+1
}
else
{de1=cbind(de1,med0$BUGSoutput$sims.list$c[,tt2])
tt2=tt2+1}
#method2: the same for binary and continuous predictors
if(p1>0){
if(y.type==1){
if(p1==1 & c1==1 & contm1[1,1]==contm1[1,2])
{ ie2[,,contm[1],l]=(med0$BUGSoutput$sims.list$alpha1.a[,1]/deltam[1])* #
(as.matrix(med0$BUGSoutput$sims.list$beta[,contm1[1,1]])%*%(M3[,contm3[1,1]]-M1[,contm1[1,1]]))
aie2[,contm[1],l]=apply(ie2[,,contm[1],l],1,mean,na.rm=T)
}
else
for (j in 1:p1){
ie2[,,contm[j],l]=(med0$BUGSoutput$sims.list$alpha1.a[,j,l]/deltam[j])* #a unit change in x, result in the unit change in m
(med0$BUGSoutput$sims.list$beta[,contm1[j,1]:contm1[j,2]]%*%t(M3[,contm3[j,1]:contm3[j,2]]-M1[,contm1[j,1]:contm1[j,2]]))
aie2[,contm[j],l]=apply(ie2[,,contm[j],l],1,mean,na.rm=T) #unit change in m, result in the unit change in y
}}
else if(y.type==2)
{if(p1==1 & c1==1 & contm1[1,1]==contm1[1,2])
{ temp.M3=M1
temp.M3[,contm1[1,1]]=M3[,contm3[1,1]]
temp.mu1=med0$BUGSoutput$sims.list$c%*%t(x)+med0$BUGSoutput$sims.list$beta%*%t(M1)+matrix(med0$BUGSoutput$sims.list$beta0,N1,nrow(x))
temp.mu3=med0$BUGSoutput$sims.list$c%*%t(x)+med0$BUGSoutput$sims.list$beta%*%t(temp.M3)+matrix(med0$BUGSoutput$sims.list$beta0,N1,nrow(x))
if(!is.null(cova))
{temp.mu1=temp.mu1+med0$BUGSoutput$sims.list$eta%*%t(cova)
temp.mu3=temp.mu3+med0$BUGSoutput$sims.list$eta%*%t(cova)}
ie2[,,contm[1],l]=(med0$BUGSoutput$sims.list$alpha1.a[,1]/deltam[1])*(exp(temp.mu3)/(1+exp(temp.mu3))-exp(temp.mu1)/(1+exp(temp.mu1)))
aie2[,contm[1],l]=apply(ie2[,,contm[1],l],1,mean,na.rm=T)
}
else
for (j in 1:p1){
temp.M3=M1
temp.M3[,contm1[j,1]:contm1[j,2]]=M3[,contm3[j,1]:contm3[j,2]]
temp.mu1=med0$BUGSoutput$sims.list$c%*%t(x)+med0$BUGSoutput$sims.list$beta%*%t(M1)+matrix(med0$BUGSoutput$sims.list$beta0,N1,nrow(x))
temp.mu3=med0$BUGSoutput$sims.list$c%*%t(x)+med0$BUGSoutput$sims.list$beta%*%t(temp.M3)+matrix(med0$BUGSoutput$sims.list$beta0,N1,nrow(x))
if(!is.null(cova))
{temp.mu1=temp.mu1+med0$BUGSoutput$sims.list$eta%*%t(cova)
temp.mu3=temp.mu3+med0$BUGSoutput$sims.list$eta%*%t(cova)}
ie2[,,contm[j],l]=(med0$BUGSoutput$sims.list$alpha1.a[,j,l]/deltam[j])*(exp(temp.mu3)/(1+exp(temp.mu3))-exp(temp.mu1)/(1+exp(temp.mu1)))
aie2[,contm[j],l]=apply(ie2[,,contm[j],l],1,mean,na.rm=T)
}}
else if(y.type==4)
{if(p1==1 & c1==1 & contm1[1,1]==contm1[1,2])
{temp.M3=M1
temp.M3[,contm1[1,1]]=M3[,contm3[1,1]]
temp.mu1=med0$BUGSoutput$sims.list$c%*%t(x)+med0$BUGSoutput$sims.list$beta%*%t(M1)
temp.mu3=med0$BUGSoutput$sims.list$c%*%t(x)+med0$BUGSoutput$sims.list$beta%*%t(temp.M3)
if(!is.null(cova))
{temp.mu1=temp.mu1+med0$BUGSoutput$sims.list$eta%*%t(cova)
temp.mu3=temp.mu3+med0$BUGSoutput$sims.list$eta%*%t(cova)}
vec=cbind(med0$BUGSoutput$sims.list$r,
as.vector(med0$BUGSoutput$sims.list$lambda^(-1/med0$BUGSoutput$sims.list$r))*(1/apply(exp(temp.mu3),2,expp, 1/med0$BUGSoutput$sims.list$r)))
tmean3=t(apply(vec,1,weib.trunc.mean,right=tmax))
#if(multi)
# tmean3=log(tmean3)
vec=cbind(med0$BUGSoutput$sims.list$r,
as.vector(med0$BUGSoutput$sims.list$lambda^(-1/med0$BUGSoutput$sims.list$r))*(1/apply(exp(temp.mu1),2,expp, 1/med0$BUGSoutput$sims.list$r)))
tmean1=t(apply(vec,1,weib.trunc.mean,right=tmax))
#if(multi)
# tmean1=log(tmean1)
#tmean3=ifelse(tmean3==Inf,tmax-runif(1,0,0.1),tmean3)
#tmean1=ifelse(tmean1==Inf,tmax-runif(1,0,0.1),tmean1)
ie2[,,contm[1],l]=as.vector(med0$BUGSoutput$sims.list$alpha1.a[,1]/deltam[1])*(tmean3-tmean1)
aie2[,contm[1],l]=apply(ie2[,,contm[1],l],1,mean,na.rm=T)
}
else
for (j in 1:p1){
temp.M3=M1
temp.M3[,contm1[j,1]:contm1[j,2]]=M3[,contm3[j,1]:contm3[j,2]]
temp.mu1=med0$BUGSoutput$sims.list$c%*%t(x)+med0$BUGSoutput$sims.list$beta%*%t(M1)
temp.mu3=med0$BUGSoutput$sims.list$c%*%t(x)+med0$BUGSoutput$sims.list$beta%*%t(temp.M3)
if(!is.null(cova))
{temp.mu1=temp.mu1+med0$BUGSoutput$sims.list$eta%*%t(cova)
temp.mu3=temp.mu3+med0$BUGSoutput$sims.list$eta%*%t(cova)}
vec=cbind(med0$BUGSoutput$sims.list$r,
as.vector(med0$BUGSoutput$sims.list$lambda^(-1/med0$BUGSoutput$sims.list$r))*(1/apply(exp(temp.mu3),2,expp, 1/med0$BUGSoutput$sims.list$r)))
tmean3=t(apply(vec,1,weib.trunc.mean,right=tmax))
#if(multi)
# tmean3=log(tmean3)
vec=cbind(med0$BUGSoutput$sims.list$r,
as.vector(med0$BUGSoutput$sims.list$lambda^(-1/med0$BUGSoutput$sims.list$r))*(1/apply(exp(temp.mu1),2,expp, 1/med0$BUGSoutput$sims.list$r)))
tmean1=t(apply(vec,1,weib.trunc.mean,right=tmax))
#if(multi)
# tmean1=log(tmean1)
#tmean3=ifelse(tmean3==Inf,tmax-runif(1,0,0.1),tmean3)
#tmean1=ifelse(tmean1==Inf,tmax-runif(1,0,0.1),tmean1)
ie2[,,contm[j],l]=as.vector(med0$BUGSoutput$sims.list$alpha1.a[,j,l]/deltam[j])*(tmean3-tmean1)
aie2[,contm[j],l]=apply(ie2[,,contm[j],l],1,mean)
}}
}
#for binary and categorical mediators, method 2 and method 1 are not the same
if(p2>0){
if(y.type==1){
if(p2==1 & c1==1){
if (nmc==1){
temp.x1=med0$BUGSoutput$sims.list$alpha0.b[,1]+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,1])%*%t(x1.temp)
temp.x3=med0$BUGSoutput$sims.list$alpha0.b[,1]+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,1])%*%t(x3.temp)
}
else
{temp.x1=med0$BUGSoutput$sims.list$alpha0.b[,1,mind[binm[1],]]%*%t(mcov[,mind[binm[1],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,1])%*%t(x1.temp)
temp.x3=med0$BUGSoutput$sims.list$alpha0.b[,1,mind[binm[1],]]%*%t(mcov[,mind[binm[1],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,1])%*%t(x3.temp)}
ie2[,,binm[1],1]=med0$BUGSoutput$sims.list$beta[,binm1[1]]*(exp(temp.x3)/(1+exp(temp.x3))-exp(temp.x1)/(1+exp(temp.x1)))/deltap[l]
aie2[,binm[1],1] <-apply(ie2[,,binm[1],1],1,mean,na.rm=T)}
else
for (k in 1:p2){
if(nmc==1 & p2==1){
temp.x1=med0$BUGSoutput$sims.list$alpha0.b[,k]+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,k,])%*%t(x1.temp)
temp.x3=med0$BUGSoutput$sims.list$alpha0.b[,k]+med0$BUGSoutput$sims.list$alpha1.b[,k,]%*%t(x3.temp)
}
else{
temp.x1=med0$BUGSoutput$sims.list$alpha0.b[,k,mind[binm[k],]]%*%t(mcov[,mind[binm[k],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,k,])%*%t(x1.temp)
temp.x3=med0$BUGSoutput$sims.list$alpha0.b[,k,mind[binm[k],]]%*%t(mcov[,mind[binm[k],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,k,])%*%t(x3.temp)
}
ie2[,,binm[k],l]=med0$BUGSoutput$sims.list$beta[,binm1[k]]*(exp(temp.x3)/(1+exp(temp.x3))-exp(temp.x1)/(1+exp(temp.x1)))/deltap[l]
aie2[,binm[k],l] <- apply(ie2[,,binm[k],l],1,mean,na.rm=T)
}}
else if(y.type==2){
if(p2==1 & c1==1){
temp.M3=M1
temp.M1=M1
temp.M3[,binm1[1]]=1
temp.M1[,binm1[1]]=0
temp.mu1=med0$BUGSoutput$sims.list$c%*%t(x)+med0$BUGSoutput$sims.list$beta%*%t(temp.M1)+matrix(med0$BUGSoutput$sims.list$beta0,N1,nrow(x))
temp.mu3=med0$BUGSoutput$sims.list$c%*%t(x)+med0$BUGSoutput$sims.list$beta%*%t(temp.M3)+matrix(med0$BUGSoutput$sims.list$beta0,N1,nrow(x))
if(!is.null(cova))
{temp.mu1=temp.mu1+med0$BUGSoutput$sims.list$eta%*%t(cova)
temp.mu3=temp.mu3+med0$BUGSoutput$sims.list$eta%*%t(cova)}
bpart=exp(temp.mu3)/(1+exp(temp.mu3))-exp(temp.mu1)/(1+exp(temp.mu1))
if (nmc==1){
temp.x1=med0$BUGSoutput$sims.list$alpha0.b[,1]+matrix(med0$BUGSoutput$sims.list$alpha1.b[,1])%*%t(x1.temp)
temp.x3=med0$BUGSoutput$sims.list$alpha0.b[,1]+matrix(med0$BUGSoutput$sims.list$alpha1.b[,1])%*%t(x3.temp)
}
else
{temp.x1=med0$BUGSoutput$sims.list$alpha0.b[,1,mind[binm[1],]]%*%t(mcov[,mind[binm[1],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,1])%*%t(x1.temp)
temp.x3=med0$BUGSoutput$sims.list$alpha0.b[,1,mind[binm[1],]]%*%t(mcov[,mind[binm[1],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,1])%*%t(x3.temp)}
ie2[,,binm[1],1]=bpart*(exp(temp.x3)/(1+exp(temp.x3))-exp(temp.x1)/(1+exp(temp.x1)))/deltap[l]
aie2[,binm[1],1] <-apply(ie2[,,binm[1],1],1,mean,na.rm=T)}
else
for (k in 1:p2){
temp.M3=M1
temp.M1=M1
temp.M3[,binm1[k]]=1
temp.M1[,binm1[k]]=0
temp.mu1=med0$BUGSoutput$sims.list$c%*%t(x)+med0$BUGSoutput$sims.list$beta%*%t(temp.M1)+matrix(med0$BUGSoutput$sims.list$beta0,N1,nrow(x))
temp.mu3=med0$BUGSoutput$sims.list$c%*%t(x)+med0$BUGSoutput$sims.list$beta%*%t(temp.M3)+matrix(med0$BUGSoutput$sims.list$beta0,N1,nrow(x))
if(!is.null(cova))
{temp.mu1=temp.mu1+med0$BUGSoutput$sims.list$eta%*%t(cova)
temp.mu3=temp.mu3+med0$BUGSoutput$sims.list$eta%*%t(cova)}
bpart=exp(temp.mu3)/(1+exp(temp.mu3))-exp(temp.mu1)/(1+exp(temp.mu1))
if(nmc==1 & p2==1){
temp.x1=med0$BUGSoutput$sims.list$alpha0.b[,k]+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,k,])%*%t(x1.temp)
temp.x3=med0$BUGSoutput$sims.list$alpha0.b[,k]+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,k,])%*%t(x3.temp)
}
else{
temp.x1=med0$BUGSoutput$sims.list$alpha0.b[,k,mind[binm[k],]]%*%t(mcov[,mind[binm[k],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,k,])%*%t(x1.temp)
temp.x3=med0$BUGSoutput$sims.list$alpha0.b[,k,mind[binm[k],]]%*%t(mcov[,mind[binm[k],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,k,])%*%t(x3.temp)
}
ie2[,,binm[k],l]=bpart*(exp(temp.x3)/(1+exp(temp.x3))-exp(temp.x1)/(1+exp(temp.x1)))/deltap[l]
aie2[,binm[k],l] <- apply(ie2[,,binm[k],l],1,mean,na.rm=T)
}}
else if(y.type==4){
if(p2==1 & c1==1){
temp.M3=M1
temp.M1=M1
temp.M3[,binm1[1]]=1
temp.M1[,binm1[1]]=0
temp.mu1=med0$BUGSoutput$sims.list$c%*%t(x)+med0$BUGSoutput$sims.list$beta%*%t(temp.M1)
temp.mu3=med0$BUGSoutput$sims.list$c%*%t(x)+med0$BUGSoutput$sims.list$beta%*%t(temp.M3)
if(!is.null(cova))
{temp.mu1=temp.mu1+med0$BUGSoutput$sims.list$eta%*%t(cova)
temp.mu3=temp.mu3+med0$BUGSoutput$sims.list$eta%*%t(cova)}
vec=cbind(med0$BUGSoutput$sims.list$r,
as.vector(med0$BUGSoutput$sims.list$lambda^(-1/med0$BUGSoutput$sims.list$r))*(1/apply(exp(temp.mu3),2,expp, 1/med0$BUGSoutput$sims.list$r)))
tmean3=t(apply(vec,1,weib.trunc.mean,right=tmax))
#if(multi)
# tmean3=log(tmean3)
vec=cbind(med0$BUGSoutput$sims.list$r,
as.vector(med0$BUGSoutput$sims.list$lambda^(-1/med0$BUGSoutput$sims.list$r))*(1/apply(exp(temp.mu1),2,expp, 1/med0$BUGSoutput$sims.list$r)))
tmean1=t(apply(vec,1,weib.trunc.mean,right=tmax))
#if(multi)
# tmean1=log(tmean1)
#tmean3=ifelse(tmean3==Inf,tmax-runif(1,0,0.1),tmean3)
#tmean1=ifelse(tmean1==Inf,tmax-runif(1,0,0.1),tmean1)
bpart=tmean3-tmean1
if (nmc==1){
temp.x1=med0$BUGSoutput$sims.list$alpha0.b[,1]+matrix(med0$BUGSoutput$sims.list$alpha1.b[,1])%*%t(x1.temp)
temp.x3=med0$BUGSoutput$sims.list$alpha0.b[,1]+matrix(med0$BUGSoutput$sims.list$alpha1.b[,1])%*%t(x3.temp)
}
else
{temp.x1=med0$BUGSoutput$sims.list$alpha0.b[,1,mind[binm[1],]]%*%t(mcov[,mind[binm[1],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,1])%*%t(x1.temp)
temp.x3=med0$BUGSoutput$sims.list$alpha0.b[,1,mind[binm[1],]]%*%t(mcov[,mind[binm[1],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,1])%*%t(x3.temp)}
ie2[,,binm[1],1]=bpart*(exp(temp.x3)/(1+exp(temp.x3))-exp(temp.x1)/(1+exp(temp.x1)))/deltap[l]
aie2[,binm[1],1] <-apply(ie2[,,binm[1],1],1,mean,na.rm=T)}
else
for (k in 1:p2){
temp.M3=M1
temp.M1=M1
temp.M3[,binm1[k]]=1
temp.M1[,binm1[k]]=0
temp.mu1=med0$BUGSoutput$sims.list$c%*%t(x)+med0$BUGSoutput$sims.list$beta%*%t(temp.M1)
temp.mu3=med0$BUGSoutput$sims.list$c%*%t(x)+med0$BUGSoutput$sims.list$beta%*%t(temp.M3)
if(!is.null(cova))
{temp.mu1=temp.mu1+med0$BUGSoutput$sims.list$eta%*%t(cova)
temp.mu3=temp.mu3+med0$BUGSoutput$sims.list$eta%*%t(cova)}
vec=cbind(med0$BUGSoutput$sims.list$r,
as.vector(med0$BUGSoutput$sims.list$lambda^(-1/med0$BUGSoutput$sims.list$r))*(1/apply(exp(temp.mu3),2,expp, 1/med0$BUGSoutput$sims.list$r)))
tmean3=t(apply(vec,1,weib.trunc.mean,right=tmax))
#if(multi)
# tmean3=log(tmean3)
vec=cbind(med0$BUGSoutput$sims.list$r,
as.vector(med0$BUGSoutput$sims.list$lambda^(-1/med0$BUGSoutput$sims.list$r))*(1/apply(exp(temp.mu1),2,expp, 1/med0$BUGSoutput$sims.list$r)))
tmean1=t(apply(vec,1,weib.trunc.mean,right=tmax))
#if(multi)
# tmean1=log(tmean1)
#tmean3=ifelse(tmean3==Inf,tmax-runif(1,0,0.1),tmean3)
#tmean1=ifelse(tmean1==Inf,tmax-runif(1,0,0.1),tmean1)
bpart=tmean3-tmean1
if(nmc==1 & p2==1){
temp.x1=med0$BUGSoutput$sims.list$alpha0.b[,k]+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,k,])%*%t(x1.temp)
temp.x3=med0$BUGSoutput$sims.list$alpha0.b[,k]+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,k,])%*%t(x3.temp)
}
else{
temp.x1=med0$BUGSoutput$sims.list$alpha0.b[,k,mind[binm[k],]]%*%t(mcov[,mind[binm[k],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,k,])%*%t(x1.temp)
temp.x3=med0$BUGSoutput$sims.list$alpha0.b[,k,mind[binm[k],]]%*%t(mcov[,mind[binm[k],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,k,])%*%t(x3.temp)
}
ie2[,,binm[k],l]=bpart*(exp(temp.x3)/(1+exp(temp.x3))-exp(temp.x1)/(1+exp(temp.x1)))/deltap[l]
aie2[,binm[k],l] <- apply(ie2[,,binm[k],l],1,mean,na.rm=T)
}}
}
if(p3>0){
if(y.type==1)
for (j in 1:p3){
mu_Mc1<-array(0,c(N1,N,cat2[j]-1))
mu_Mc0<-array(0,c(N1,N,cat2[j]-1))
for (k in 2:cat2[j]){
mu_Mc1[,,k-1] <- exp(as.matrix(med0$BUGSoutput$sims.list$alpha0.c[,j,k-1,mind[catm[j],]])%*%t(mcov[,mind[catm[j],]])+med0$BUGSoutput$sims.list$alpha1.c[,j,k-1,]%*%t(x3.temp))
mu_Mc0[,,k-1] <- exp(as.matrix(med0$BUGSoutput$sims.list$alpha0.c[,j,k-1,mind[catm[j],]])%*%t(mcov[,mind[catm[j],]])+med0$BUGSoutput$sims.list$alpha1.c[,j,k-1,]%*%t(x1.temp))
}
sum_Mc1 <-apply(mu_Mc1,c(1,2),sum)+1
sum_Mc0 <-apply(mu_Mc0,c(1,2),sum)+1
for (k in 2:cat2[j])
ie2[,,catm[j],l]=ie2[,,catm[j],l]+(mu_Mc1[,,k-1]/sum_Mc1-mu_Mc0[,,k-1]/sum_Mc0)*med0$BUGSoutput$sims.list$beta[,catm1[j,1]+k-2]/deltap[l]
aie2[,catm[j],l]<-apply(ie2[,,catm[j],l],1,mean,na.rm=T)
}
else if(y.type==2)
for (j in 1:p3){
bpart=array(0,c(N1,N,cat2[j]-1))
M1.temp=M1
M1.temp[,catm1[j,1]:catm1[j,2]]=0
for (k in catm1[j,1]:catm1[j,2]){
temp.M3=M1.temp
temp.M1=M1.temp
temp.M3[,k]=1
temp.mu1=med0$BUGSoutput$sims.list$c%*%t(x)+med0$BUGSoutput$sims.list$beta%*%t(temp.M1)+matrix(med0$BUGSoutput$sims.list$beta0,N1,nrow(x))
temp.mu3=med0$BUGSoutput$sims.list$c%*%t(x)+med0$BUGSoutput$sims.list$beta%*%t(temp.M3)+matrix(med0$BUGSoutput$sims.list$beta0,N1,nrow(x))
if(!is.null(cova))
{temp.mu1=temp.mu1+med0$BUGSoutput$sims.list$eta%*%t(cova)
temp.mu3=temp.mu3+med0$BUGSoutput$sims.list$eta%*%t(cova)}
bpart[,,k-catm1[j,1]+1]=exp(temp.mu3)/(1+exp(temp.mu3))-exp(temp.mu1)/(1+exp(temp.mu1))
}
mu_Mc1<-array(0,c(N1,N,cat2[j]-1))
mu_Mc0<-array(0,c(N1,N,cat2[j]-1))
for (k in 2:cat2[j]){
mu_Mc1[,,k-1] <- exp(as.matrix(med0$BUGSoutput$sims.list$alpha0.c[,j,k-1,mind[catm[j],]])%*%t(mcov[,mind[catm[j],]])+med0$BUGSoutput$sims.list$alpha1.c[,j,k-1,]%*%t(x3.temp))
mu_Mc0[,,k-1] <- exp(as.matrix(med0$BUGSoutput$sims.list$alpha0.c[,j,k-1,mind[catm[j],]])%*%t(mcov[,mind[catm[j],]])+med0$BUGSoutput$sims.list$alpha1.c[,j,k-1,]%*%t(x1.temp))
}
sum_Mc1 <-apply(mu_Mc1,c(1,2),sum)+1
sum_Mc0 <-apply(mu_Mc0,c(1,2),sum)+1
for (k in 2:cat2[j])
ie2[,,catm[j],l]=ie2[,,catm[j],l]+(mu_Mc1[,,k-1]/sum_Mc1-mu_Mc0[,,k-1]/sum_Mc0)*bpart[,,k-1]
# for (k in 1:N)
# ie2[,k,catm[j],l]=diag((diag(1/sum_Mc1[,k])%*%mu_Mc1[,k,]-diag(1/sum_Mc0[,k])%*%mu_Mc0[,k,])%*%t(bpart[,k,]))
aie2[,catm[j],l]<-apply(ie2[,,catm[j],l],1,mean,na.rm=T)
}
else if(y.type==4)
for (j in 1:p3){
bpart=array(0,c(N1,N,cat2[j]-1))
M1.temp=M1
M1.temp[,catm1[j,1]:catm1[j,2]]=0
for (k in catm1[j,1]:catm1[j,2]){
temp.M3=M1.temp
temp.M1=M1.temp
temp.M3[,k]=1
temp.mu1=med0$BUGSoutput$sims.list$c%*%t(x)+med0$BUGSoutput$sims.list$beta%*%t(temp.M1)
temp.mu3=med0$BUGSoutput$sims.list$c%*%t(x)+med0$BUGSoutput$sims.list$beta%*%t(temp.M3)
if(!is.null(cova))
{temp.mu1=temp.mu1+med0$BUGSoutput$sims.list$eta%*%t(cova)
temp.mu3=temp.mu3+med0$BUGSoutput$sims.list$eta%*%t(cova)}
vec=cbind(med0$BUGSoutput$sims.list$r,
as.vector(med0$BUGSoutput$sims.list$lambda^(-1/med0$BUGSoutput$sims.list$r))*(1/apply(exp(temp.mu3),2,expp, 1/med0$BUGSoutput$sims.list$r)))
tmean3=t(apply(vec,1,weib.trunc.mean,right=tmax))
#if(multi)
# tmean3=log(tmean3)
vec=cbind(med0$BUGSoutput$sims.list$r,
as.vector(med0$BUGSoutput$sims.list$lambda^(-1/med0$BUGSoutput$sims.list$r))*(1/apply(exp(temp.mu1),2,expp, 1/med0$BUGSoutput$sims.list$r)))
tmean1=t(apply(vec,1,weib.trunc.mean,right=tmax))
#if(multi)
# tmean1=log(tmean1)
#tmean3=ifelse(tmean3==Inf,tmax-runif(1,0,0.1),tmean3)
#tmean1=ifelse(tmean1==Inf,tmax-runif(1,0,0.1),tmean1)
bpart[,,k-catm1[j,1]+1]=tmean3-tmean1
# bpart[,,k-catm1[j,1]+1]=as.vector(gamma(1+1/med0$BUGSoutput$sims.list$r)/med0$BUGSoutput$sims.list$lambda^(1/med0$BUGSoutput$sims.list$r))*
# (1/apply(exp(temp.mu3),2,expp, 1/med0$BUGSoutput$sims.list$r)-1/apply(exp(temp.mu1),2,expp,1/med0$BUGSoutput$sims.list$r))
}
mu_Mc1<-array(0,c(N1,N,cat2[j]-1))
mu_Mc0<-array(0,c(N1,N,cat2[j]-1))
for (k in 2:cat2[j]){
mu_Mc1[,,k-1] <- exp(as.matrix(med0$BUGSoutput$sims.list$alpha0.c[,j,k-1,mind[catm[j],]])%*%t(mcov[,mind[catm[j],]])+med0$BUGSoutput$sims.list$alpha1.c[,j,k-1,]%*%t(x3.temp))
mu_Mc0[,,k-1] <- exp(as.matrix(med0$BUGSoutput$sims.list$alpha0.c[,j,k-1,mind[catm[j],]])%*%t(mcov[,mind[catm[j],]])+med0$BUGSoutput$sims.list$alpha1.c[,j,k-1,]%*%t(x1.temp))
}
sum_Mc1 <-apply(mu_Mc1,c(1,2),sum)+1
sum_Mc0 <-apply(mu_Mc0,c(1,2),sum)+1
for (k in 2:cat2[j])
ie2[,,catm[j],l]=ie2[,,catm[j],l]+(mu_Mc1[,,k-1]/sum_Mc1-mu_Mc0[,,k-1]/sum_Mc0)*bpart[,,k-1]
aie2[,catm[j],l]<-apply(ie2[,,catm[j],l],1,mean,na.rm=T)
}
}
if(y.type==2)
{temp.x1=x
temp.x3=x
if(l%in%data0$contpred1.0)
{tt1=match(l,data0$contpred1.0)
temp.x3[,data0$contpred2[tt1,1]:data0$contpred2[tt1,2]]=data0$pred3[,data0$contpred3[tt1,1]:data0$contpred3[tt1,2]]}
else if(l%in%data0$binpred1.0)
{tt1=match(l,data0$binpred1)
temp.x1[,data0$binpred2[tt1]]=0
temp.x3[,data0$binpred2[tt1]]=1
deltap[l]=1}
else
{for (i in 1:nrow(data0$catpred1.0))
if(l%in%(data0$catpred1.0[i,1]:data0$catpred1.0[i,2]))
tt1=i
temp.x1[,data0$catpred2[tt1,1]:data0$catpred2[tt1,2]]=0
temp.x3[,data0$catpred2[tt1,1]:data0$catpred2[tt1,2]]=0
temp.x3[,l]=1
deltap[l]=1}
temp.mu1=med0$BUGSoutput$sims.list$c%*%t(temp.x1)+med0$BUGSoutput$sims.list$beta%*%t(M1)+matrix(med0$BUGSoutput$sims.list$beta0,N1,nrow(x))
temp.mu3=med0$BUGSoutput$sims.list$c%*%t(temp.x3)+med0$BUGSoutput$sims.list$beta%*%t(M1)+matrix(med0$BUGSoutput$sims.list$beta0,N1,nrow(x))
if(!is.null(cova))
{temp.mu1=temp.mu1+med0$BUGSoutput$sims.list$eta%*%t(cova)
temp.mu3=temp.mu3+med0$BUGSoutput$sims.list$eta%*%t(cova)}
de2.1=(exp(temp.mu3)/(1+exp(temp.mu3))-exp(temp.mu1)/(1+exp(temp.mu1)))/deltap[l]
de2=cbind(de2,apply(de2.1,1,mean,na.rm=T))}
else if(y.type==4)
{temp.x1=x
temp.x3=x
if(l%in%data0$contpred1.0)
{tt1=match(l,data0$contpred1.0)
temp.x3[,data0$contpred2[tt1,1]:data0$contpred2[tt1,2]]=data0$pred3[,data0$contpred3[tt1,1]:data0$contpred3[tt1,2]]}
else if(l%in%data0$binpred1.0)
{tt1=match(l,data0$binpred1)
temp.x1[,data0$binpred2[tt1]]=0
temp.x3[,data0$binpred2[tt1]]=1
deltap[l]=1}
else
{for (i in 1:nrow(data0$catpred1.0))
if(l%in%(data0$catpred1.0[i,1]:data0$catpred1.0[i,2]))
tt1=i
temp.x1[,data0$catpred2[tt1,1]:data0$catpred2[tt1,2]]=0
temp.x3[,data0$catpred2[tt1,1]:data0$catpred2[tt1,2]]=0
temp.x3[,l]=1
deltap[l]=1}
temp.mu1=med0$BUGSoutput$sims.list$c%*%t(temp.x1)+med0$BUGSoutput$sims.list$beta%*%t(M1)
temp.mu3=med0$BUGSoutput$sims.list$c%*%t(temp.x3)+med0$BUGSoutput$sims.list$beta%*%t(M1)
if(!is.null(cova))
{temp.mu1=temp.mu1+med0$BUGSoutput$sims.list$eta%*%t(cova)
temp.mu3=temp.mu3+med0$BUGSoutput$sims.list$eta%*%t(cova)}
vec=cbind(med0$BUGSoutput$sims.list$r,
as.vector(med0$BUGSoutput$sims.list$lambda^(-1/med0$BUGSoutput$sims.list$r))*(1/apply(exp(temp.mu3),2,expp, 1/med0$BUGSoutput$sims.list$r)))
tmean3=t(apply(vec,1,weib.trunc.mean,right=tmax))
#if(multi)
# tmean3=log(tmean3)
vec=cbind(med0$BUGSoutput$sims.list$r,
as.vector(med0$BUGSoutput$sims.list$lambda^(-1/med0$BUGSoutput$sims.list$r))*(1/apply(exp(temp.mu1),2,expp, 1/med0$BUGSoutput$sims.list$r)))
tmean1=t(apply(vec,1,weib.trunc.mean,right=tmax))
#if(multi)
# tmean1=log(tmean1)
de2.1=(tmean3-tmean1)/deltap[l]
de2=cbind(de2,apply(de2.1,1,mean,na.rm=T))}
#method 3:parametric
#3.1. get M1(x) and M1(x+dx) for y
if(p1>0){
if(c1==1 & p1+p2+p3==1 & contm1[1,1]==contm1[1,2]){
if(nmc==1)
{mu_M2[,contm1[1,1],] <- med0$BUGSoutput$sims.list$alpha0[,contm1[1,1]]+as.matrix(med0$BUGSoutput$sims.list$alpha1[,contm1[1,1]])%*%t(x1.temp)
mu_M3[,contm1[1,1],] <- med0$BUGSoutput$sims.list$alpha0[,contm1[1,1]]+as.matrix(med0$BUGSoutput$sims.list$alpha1[,contm1[1,1]])%*%t(x3.temp)}
else
{mu_M2[,contm1[1,1],] <- med0$BUGSoutput$sims.list$alpha0[,contm1[1,1],mind[contm[1],]]%*%t(mcov[,mind[contm[1],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1[,contm1[1,1]])%*%t(x1.temp)
mu_M3[,contm1[1,1],] <- med0$BUGSoutput$sims.list$alpha0[,contm1[1,1],mind[contm[1],]]%*%t(mcov[,mind[contm[1],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1[,contm1[1,1]])%*%t(x3.temp)}
}
else
for (j in 1:p1){
for (k in contm1[j,1]:contm1[j,2]){
if(nmc==1 & p1+p2+p3==1)
{mu_M2[,k,] <- med0$BUGSoutput$sims.list$alpha0[,k]+med0$BUGSoutput$sims.list$alpha1[,k,l]%*%t(x1.temp)
mu_M3[,k,] <- med0$BUGSoutput$sims.list$alpha0[,k]+med0$BUGSoutput$sims.list$alpha1[,k,l]%*%t(x3.temp)}
else
{mu_M2[,k,] <- med0$BUGSoutput$sims.list$alpha0[,k,mind[contm[j],]]%*%t(mcov[,mind[contm[j],]])+med0$BUGSoutput$sims.list$alpha1[,k,]%*%t(x1.temp)
mu_M3[,k,] <- med0$BUGSoutput$sims.list$alpha0[,k,mind[contm[j],]]%*%t(mcov[,mind[contm[j],]])+med0$BUGSoutput$sims.list$alpha1[,k,]%*%t(x3.temp)}
}}}
if(p2>0){
if(p2==1 & c1==1)
{if(nmc==1)
{temp.x1=med0$BUGSoutput$sims.list$alpha0.b[,1]+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,1])%*%t(x1.temp)
temp.x3=med0$BUGSoutput$sims.list$alpha0.b[,1]+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,1])%*%t(x3.temp)}
else
{temp.x1=med0$BUGSoutput$sims.list$alpha0.b[,1,mind[binm[1],]]%*%t(mcov[,mind[binm[1],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,1])%*%t(x1.temp)
temp.x3=med0$BUGSoutput$sims.list$alpha0.b[,1,mind[binm[1],]]%*%t(mcov[,mind[binm[1],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,1])%*%t(x3.temp)}
mu_M2[,binm[1],] <- exp(temp.x1)/(1+exp(temp.x1))
mu_M3[,binm[1],] <- exp(temp.x3)/(1+exp(temp.x3))
}
else{
for (k in 1:p2){
if(nmc==1 & p2==1)
{temp.x1=med0$BUGSoutput$sims.list$alpha0.b[,k]+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,k,])%*%t(x1.temp)
temp.x3=med0$BUGSoutput$sims.list$alpha0.b[,k]+med0$BUGSoutput$sims.list$alpha1.b[,k,]%*%t(x3.temp)}
else
{temp.x1=med0$BUGSoutput$sims.list$alpha0.b[,k,mind[binm[k],]]%*%t(mcov[,mind[binm[k],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,k,])%*%t(x1.temp)
temp.x3=med0$BUGSoutput$sims.list$alpha0.b[,k,mind[binm[k],]]%*%t(mcov[,mind[binm[k],]])+as.matrix(med0$BUGSoutput$sims.list$alpha1.b[,k,])%*%t(x3.temp)}
mu_M2[,binm1[k],] <- exp(temp.x1)/(1+exp(temp.x1))
mu_M3[,binm1[k],] <- exp(temp.x3)/(1+exp(temp.x3))}
}}
if(p3>0){
for (j in 1:p3){
mu_Mc1<-array(0,c(N1,N,cat2[j]-1))
mu_Mc0<-array(0,c(N1,N,cat2[j]-1))
for (k in 2:cat2[j]){
mu_Mc1[,,k-1] <- exp(as.matrix(med0$BUGSoutput$sims.list$alpha0.c[,j,k-1,mind[catm[j],]])%*%t(mcov[,mind[catm[j],]])+med0$BUGSoutput$sims.list$alpha1.c[,j,k-1,]%*%t(x3.temp))
mu_Mc0[,,k-1] <- exp(as.matrix(med0$BUGSoutput$sims.list$alpha0.c[,j,k-1,mind[catm[j],]])%*%t(mcov[,mind[catm[j],]])+med0$BUGSoutput$sims.list$alpha1.c[,j,k-1,]%*%t(x1.temp))
}
sum_Mc1 <-apply(mu_Mc1,c(1,2),sum)+1
sum_Mc0 <-apply(mu_Mc0,c(1,2),sum)+1
for (k in 2:cat2[j])
{mu_M2[,catm1[j,1]+k-2,]=mu_Mc0[,,k-1]/sum_Mc0
mu_M3[,catm1[j,1]+k-2,]=mu_Mc1[,,k-1]/sum_Mc1}
}}
#3.2. get x and dx for y
x1.temp1=x
x3.temp1=x
if(l%in%as.vector(data0$contpred1.0)){ #need the continuous predictor in its original format
i=match(l,data0$contpred1.0)
x3.temp1[,data0$contpred2[i,1]:data0$contpred2[i,2]]=data0$pred3[,data0$contpred3[i,1]:data0$contpred3[i,2]]
}
else if(l%in%data0$binpred1.0)
{i=match(l,data0$binpred1.0)
x1.temp1[,data0$binpred2[i]]=0
x3.temp1[,data0$binpred2[i]]=1
deltap[l]=1}
else{ #categorical predictor
for (i in 1:nrow(data0$catpred1.0))
if(l%in%(data0$catpred1.0[i,1]:data0$catpred1.0[i,2]))
{x1.temp1[,data0$catpred2[i,1]:data0$catpred2[i,2]]=0
x3.temp1[,data0$catpred2[i,1]:data0$catpred2[i,2]]=0
di=match(l,data0$catpred1.0[i,1]:data0$catpred1.0[i,2])
x3.temp1[,data0$catpred2[i,1]+di-1]=1}
deltap[l]=1
}
#3.3. get the total effect
temp1=array(0,c(nrow(M1)+1,ncol(M1),N1))
temp1[2:(nrow(M1)+1),,]=mu_M2
temp1[1,,]=t(med0$BUGSoutput$sims.list$beta)
temp2=temp1
temp2[2:(nrow(M1)+1),,]=mu_M3
if(is.null(cova))
{mu_y0<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x1.temp1) + t(apply(temp1,3,matrix.prod))
mu_y1<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x3.temp1) + t(apply(temp2,3,matrix.prod))
}
else
{mu_y0<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x1.temp1) + med0$BUGSoutput$sims.list$eta%*%t(as.matrix(cova))+t(apply(temp1,3,matrix.prod))
mu_y1<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x3.temp1) + med0$BUGSoutput$sims.list$eta%*%t(as.matrix(cova))+t(apply(temp2,3,matrix.prod))
} #get the linear part
if(y.type==2)
te3=(exp(mu_y1)/(1+exp(mu_y1))-exp(mu_y0)/(1+exp(mu_y0)))/deltap[l]
else if(y.type==1)
te3<- (mu_y1-mu_y0)/deltap[l]
else if (y.type==4)
{mu_y0<- mu_y0-matrix(med0$BUGSoutput$sims.list$beta0,N1,N)
mu_y1<- mu_y1-matrix(med0$BUGSoutput$sims.list$beta0,N1,N)
vec=cbind(med0$BUGSoutput$sims.list$r,
as.vector(med0$BUGSoutput$sims.list$lambda^(-1/med0$BUGSoutput$sims.list$r))*(1/apply(exp(mu_y1),2,expp, 1/med0$BUGSoutput$sims.list$r)))
tmean3=t(apply(vec,1,weib.trunc.mean,right=tmax))
vec=cbind(med0$BUGSoutput$sims.list$r,
as.vector(med0$BUGSoutput$sims.list$lambda^(-1/med0$BUGSoutput$sims.list$r))*(1/apply(exp(mu_y0),2,expp, 1/med0$BUGSoutput$sims.list$r)))
tmean1=t(apply(vec,1,weib.trunc.mean,right=tmax))
if(multi)
{tmean1=log(tmean1)
tmean3=log(tmean3)}
te3=(tmean3-tmean1)/deltap[l]
}
ate3[,l]=apply(te3,1,mean,na.rm=T)
if(y.type==4){
if(multi)
omu3[,l]=apply(exp(tmean1),1,mean,na.rm=T)
else
omu3[,l]=apply(tmean1,1,mean,na.rm=T)}
#3.4. calculate the ie
j1=sample(1:N,size=N*N1,replace=T)
j2=sample(1:N,size=N*N1,replace=T)
#3.4.1. continuous mediators
if(p1>0){
for (j in 1:p1){
temp1.1=temp1
temp1.2=temp2
for (i in contm1[j,1]:contm1[j,2])
{temp1.1[2:(nrow(M1)+1),i,]=matrix(M1[j1,i],N,N1)
temp1.2[2:(nrow(M1)+1),i,]=matrix(M1[j1,i],N,N1)} #j2/j1
if(is.null(cova))
{mu_y0.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x1.temp1) +t(apply(temp1.1,3,matrix.prod))
mu_y1.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x3.temp1) +t(apply(temp1.2,3,matrix.prod)) }
else{
mu_y0.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x1.temp1) + med0$BUGSoutput$sims.list$eta%*%t(cova)+t(apply(temp1.1,3,matrix.prod))
mu_y1.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x3.temp1) + med0$BUGSoutput$sims.list$eta%*%t(cova)+t(apply(temp1.2,3,matrix.prod))}
if(y.type==1)
ie3[,,contm[j],l]=te3-(mu_y1.2-mu_y0.2)/deltap[l]
else if(y.type==2)
ie3[,,contm[j],l]=te3-(exp(mu_y1.2)/(1+exp(mu_y1.2))-exp(mu_y0.2)/(1+exp(mu_y0.2)))/deltap[l]
else if (y.type==4)
{mu_y0.2<- mu_y0.2-matrix(med0$BUGSoutput$sims.list$beta0,N1,N)
mu_y1.2<- mu_y1.2-matrix(med0$BUGSoutput$sims.list$beta0,N1,N)
vec=cbind(med0$BUGSoutput$sims.list$r,
as.vector(med0$BUGSoutput$sims.list$lambda^(-1/med0$BUGSoutput$sims.list$r))*(1/apply(exp(mu_y1.2),2,expp, 1/med0$BUGSoutput$sims.list$r)))
tmean3=t(apply(vec,1,weib.trunc.mean,right=tmax))
if(multi)
tmean3=log(tmean3)
vec=cbind(med0$BUGSoutput$sims.list$r,
as.vector(med0$BUGSoutput$sims.list$lambda^(-1/med0$BUGSoutput$sims.list$r))*(1/apply(exp(mu_y0.2),2,expp, 1/med0$BUGSoutput$sims.list$r)))
tmean1=t(apply(vec,1,weib.trunc.mean,right=tmax))
if(multi)
tmean1=log(tmean1)
ie3[,,contm[j],l]<-te3-(tmean3-tmean1)/deltap[l]
}
}}
#3.4.2. binary mediators
if(p2>0){
for (k in 1:p2){
temp1.1=temp1
temp1.2=temp2
temp1.1[2:(nrow(M1)+1),binm1[k],]=matrix(M1[j1,binm1[k]],N,N1)
temp1.2[2:(nrow(M1)+1),binm1[k],]=matrix(M1[j1,binm1[k]],N,N1) #j2/j1
if (is.null(cova)){
mu_y0.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x1.temp1) + t(apply(temp1.1,3,matrix.prod))
mu_y1.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x3.temp1) + t(apply(temp1.2,3,matrix.prod))}
else{
mu_y0.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x1.temp1) + med0$BUGSoutput$sims.list$eta%*%t(cova)+t(apply(temp1.1,3,matrix.prod))
mu_y1.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x3.temp1) + med0$BUGSoutput$sims.list$eta%*%t(cova)+t(apply(temp1.2,3,matrix.prod))}
if(y.type==1)
ie3[,,binm[k],l]=te3-(mu_y1.2-mu_y0.2)/deltap[l]
else if(y.type==2)
ie3[,,binm[k],l]=te3-(exp(mu_y1.2)/(1+exp(mu_y1.2))-exp(mu_y0.2)/(1+exp(mu_y0.2)))/deltap[l]
else if (y.type==4)
{mu_y0.2<- mu_y0.2-matrix(med0$BUGSoutput$sims.list$beta0,N1,N)
mu_y1.2<- mu_y1.2-matrix(med0$BUGSoutput$sims.list$beta0,N1,N)
vec=cbind(med0$BUGSoutput$sims.list$r,
as.vector(med0$BUGSoutput$sims.list$lambda^(-1/med0$BUGSoutput$sims.list$r))*(1/apply(exp(mu_y1.2),2,expp, 1/med0$BUGSoutput$sims.list$r)))
tmean3=t(apply(vec,1,weib.trunc.mean,right=tmax))
if(multi)
tmean3=log(tmean3)
vec=cbind(med0$BUGSoutput$sims.list$r,
as.vector(med0$BUGSoutput$sims.list$lambda^(-1/med0$BUGSoutput$sims.list$r))*(1/apply(exp(mu_y0.2),2,expp, 1/med0$BUGSoutput$sims.list$r)))
tmean1=t(apply(vec,1,weib.trunc.mean,right=tmax))
if(multi)
tmean1=log(tmean1)
ie3[,,binm[k],l]<-te3-(tmean3-tmean1)/deltap[l]
}
}}
if(p3>0){
for (j in 1:p3){
temp1.1=temp1
temp1.2=temp2
for (i in catm1[j,1]:catm1[j,2])
{temp1.1[2:(nrow(M1)+1),i,]=matrix(M1[j1,i],N,N1)
temp1.2[2:(nrow(M1)+1),i,]=matrix(M1[j1,i],N,N1)} #j2/j1
if(is.null(cova)){
mu_y0.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x1.temp1) + t(apply(temp1.1,3,matrix.prod))
mu_y1.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x3.temp1) + t(apply(temp1.2,3,matrix.prod))}
else{
mu_y0.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x1.temp1) + med0$BUGSoutput$sims.list$eta%*%t(cova)+t(apply(temp1.1,3,matrix.prod))
mu_y1.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x3.temp1) + med0$BUGSoutput$sims.list$eta%*%t(cova)+t(apply(temp1.2,3,matrix.prod))}
if(y.type==1)
ie3[,,catm[j],l]=te3-(mu_y1.2-mu_y0.2)/deltap[l]
else if(y.type==2)
ie3[,,catm[j],l]=te3-(exp(mu_y1.2)/(1+exp(mu_y1.2))-exp(mu_y0.2)/(1+exp(mu_y0.2)))/deltap[l]
else if (y.type==4)
{mu_y0.2<- mu_y0.2-matrix(med0$BUGSoutput$sims.list$beta0,N1,N)
mu_y1.2<- mu_y1.2-matrix(med0$BUGSoutput$sims.list$beta0,N1,N)
vec=cbind(med0$BUGSoutput$sims.list$r,
as.vector(med0$BUGSoutput$sims.list$lambda^(-1/med0$BUGSoutput$sims.list$r))*(1/apply(exp(mu_y1.2),2,expp, 1/med0$BUGSoutput$sims.list$r)))
tmean3=t(apply(vec,1,weib.trunc.mean,right=tmax))
if(multi)
tmean3=log(tmean3)
vec=cbind(med0$BUGSoutput$sims.list$r,
as.vector(med0$BUGSoutput$sims.list$lambda^(-1/med0$BUGSoutput$sims.list$r))*(1/apply(exp(mu_y0.2),2,expp, 1/med0$BUGSoutput$sims.list$r)))
tmean1=t(apply(vec,1,weib.trunc.mean,right=tmax))
if(multi)
tmean1=log(tmean1)
ie3[,,catm[j],l]<-te3-(tmean3-tmean1)/deltap[l]
}
}}
aie3[,,l]<-apply(array(ie3[,,,l],c(N1,N,p1+p2+p3)),c(1,3),mean,na.rm=T)
#3.5. Calculate the de
temp1.1=temp1
temp1.2=temp2
for (i in 1:ncol(M1))
{temp1.1[2:(nrow(M1)+1),i,]=matrix(M1[j1,i],N,N1)
temp1.2[2:(nrow(M1)+1),i,]=matrix(M1[j1,i],N,N1)} #j2/j1
if(is.null(cova)){
mu_y0.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x1.temp1) + t(apply(temp1.1,3,matrix.prod))
mu_y1.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x3.temp1) + t(apply(temp1.2,3,matrix.prod))}
else{
mu_y0.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x1.temp1) + med0$BUGSoutput$sims.list$eta%*%t(cova)+t(apply(temp1.1,3,matrix.prod))
mu_y1.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x3.temp1) + med0$BUGSoutput$sims.list$eta%*%t(cova)+t(apply(temp1.2,3,matrix.prod))}
if(y.type==1)
de3=(mu_y1.2-mu_y0.2)/deltap[l]
else if(y.type==2)
de3=(exp(mu_y1.2)/(1+exp(mu_y1.2))-exp(mu_y0.2)/(1+exp(mu_y0.2)))/deltap[l]
else if (y.type==4)
{mu_y0.2<- mu_y0.2-matrix(med0$BUGSoutput$sims.list$beta0,N1,N)
mu_y1.2<- mu_y1.2-matrix(med0$BUGSoutput$sims.list$beta0,N1,N)
vec=cbind(med0$BUGSoutput$sims.list$r,
as.vector(med0$BUGSoutput$sims.list$lambda^(-1/med0$BUGSoutput$sims.list$r))*(1/apply(exp(mu_y1.2),2,expp, 1/med0$BUGSoutput$sims.list$r)))
tmean3=t(apply(vec,1,weib.trunc.mean,right=tmax))
if(multi)
tmean3=log(tmean3)
vec=cbind(med0$BUGSoutput$sims.list$r,
as.vector(med0$BUGSoutput$sims.list$lambda^(-1/med0$BUGSoutput$sims.list$r))*(1/apply(exp(mu_y0.2),2,expp, 1/med0$BUGSoutput$sims.list$r)))
tmean1=t(apply(vec,1,weib.trunc.mean,right=tmax))
if(multi)
tmean1=log(tmean1)
de3=(tmean3-tmean1)/deltap[l]
}
ade3[,l]=apply(de3,1,mean,na.rm=T)
#method3: semi-parametric for binary or categorical predictors
if(!l%in%data0$contpred1.0){
if(!is.null(data0$binpred1.0))
{if (l%*%data0$binpred1.0)
{M.0=data.frame(M1[x1[,l]==0,])
y.0=y[x1[,l]==0]}
else
{for(i in 1:nrow(data0$catpred1.0))
if(l%in%(data0$catpred1.0[i,1]:data0$catpred1.0[i,2]))
tt1=i
M.0=data.frame(M1[apply(x1[,data0$catpred1.0[tt1,1]:data0$catpred1.0[[tt1,2]]]==1,1,sum)==0,])
y.0=y[apply(x1[,data0$catpred1.0[tt1,1]:data0$catpred1.0[[tt1,2]]]==1,1,sum)==0]}}
else
{for(i in 1:nrow(data0$catpred1.0))
if(l%in%(data0$catpred1.0[i,1]:data0$catpred1.0[i,2]))
tt1=i
M.0=data.frame(M1[apply(x1[,data0$catpred1.0[tt1,1]:data0$catpred1.0[[tt1,2]]]==1,1,sum)==0,])
y.0=y[apply(x1[,data0$catpred1.0[tt1,1]:data0$catpred1.0[[tt1,2]]]==1,1,sum)==0]}
y.1=y[x1[,l]==1]
M.1=data.frame(M1[x1[,l]==1,])
j1=sample(1:N,size=N*N1,replace=T)
j2=sample(1:N,size=N*N1,replace=T)
n3=nrow(M.0)
n4=nrow(M.1)
j3=sample(1:n3,size=N*N1,replace = T)
j4=sample(1:n4,size=N*N1,replace = T)
for (i in 1:ncol(M1))
{mu.M0[,i,]=matrix(M.0[j3,i],N,N1)
mu.M1[,i,]=matrix(M.1[j4,i],N,N1)}
#4.1. get the total effect
mu_y0=matrix(y.0[j3],N1,N)
mu_y1=matrix(y.1[j4],N1,N)
temp1=array(0,c(nrow(M1)+1,ncol(M1),N1))
temp1[2:(nrow(M1)+1),,]=mu.M0
temp1[1,,]=t(med0$BUGSoutput$sims.list$beta)
temp2=temp1
temp2[2:(nrow(M1)+1),,]=mu.M1
if(is.null(cova))
{mu_y0.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x1.temp1) +t(apply(temp1,3,matrix.prod))
mu_y1.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x3.temp1) +t(apply(temp2,3,matrix.prod)) }
else{
mu_y0.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x1.temp1) + med0$BUGSoutput$sims.list$eta%*%t(cova)+t(apply(temp1,3,matrix.prod))
mu_y1.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x3.temp1) + med0$BUGSoutput$sims.list$eta%*%t(cova)+t(apply(temp2,3,matrix.prod))}
if(y.type==1)
te4<- (mu_y1.2-mu_y0.2)/deltap[l]
else if(y.type==2)
te4=(exp(mu_y1.2)/(1+exp(mu_y1.2))-exp(mu_y0.2)/(1+exp(mu_y0.2)))/deltap[l]
else if (y.type==4)
{mu_y0.2<- mu_y0.2-matrix(med0$BUGSoutput$sims.list$beta0,N1,N)
mu_y1.2<- mu_y1.2-matrix(med0$BUGSoutput$sims.list$beta0,N1,N)
vec=cbind(med0$BUGSoutput$sims.list$r,
as.vector(med0$BUGSoutput$sims.list$lambda^(-1/med0$BUGSoutput$sims.list$r))*(1/apply(exp(mu_y1.2),2,expp, 1/med0$BUGSoutput$sims.list$r)))
tmean3=t(apply(vec,1,weib.trunc.mean,right=tmax))
vec=cbind(med0$BUGSoutput$sims.list$r,
as.vector(med0$BUGSoutput$sims.list$lambda^(-1/med0$BUGSoutput$sims.list$r))*(1/apply(exp(mu_y0.2),2,expp, 1/med0$BUGSoutput$sims.list$r)))
tmean1=t(apply(vec,1,weib.trunc.mean,right=tmax))
if(multi)
{tmean1=log(tmean1)
tmean3=log(tmean3)}
te4=(tmean3-tmean1)/deltap[l]
}
ate4[,l]=apply(te4,1,mean,na.rm=T)
if(y.type==4){
if(multi)
omu4[,l]=apply(exp(tmean1),1,mean,na.rm=T)
else
omu4[,l]=apply(tmean1,1,mean,na.rm=T)}
#4.2. Get the ies
#ie for continuous mediators
if(p1>0){
for (j in 1:p1){
temp1.1=temp1
temp1.2=temp2
for (i in contm1[j,1]:contm1[j,2])
{temp1.1[2:(nrow(M1)+1),i,]=matrix(M1[j1,i],N,N1)
temp1.2[2:(nrow(M1)+1),i,]=matrix(M1[j1,i],N,N1)} #j2/j1
if(is.null(cova))
{mu_y0.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x1.temp1) +t(apply(temp1.1,3,matrix.prod))
mu_y1.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x3.temp1) +t(apply(temp1.2,3,matrix.prod)) }
else{
mu_y0.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x1.temp1) + med0$BUGSoutput$sims.list$eta%*%t(cova)+t(apply(temp1.1,3,matrix.prod))
mu_y1.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x3.temp1) + med0$BUGSoutput$sims.list$eta%*%t(cova)+t(apply(temp1.2,3,matrix.prod))}
if(y.type==1)
ie4[,,contm[j],l]=te4-(mu_y1.2-mu_y0.2)/deltap[l]
else if(y.type==2)
ie4[,,contm[j],l]=te4-(exp(mu_y1.2)/(1+exp(mu_y1.2))-exp(mu_y0.2)/(1+exp(mu_y0.2)))/deltap[l]
else if (y.type==4)
{mu_y0.2<- mu_y0.2-matrix(med0$BUGSoutput$sims.list$beta0,N1,N)
mu_y1.2<- mu_y1.2-matrix(med0$BUGSoutput$sims.list$beta0,N1,N)
vec=cbind(med0$BUGSoutput$sims.list$r,
as.vector(med0$BUGSoutput$sims.list$lambda^(-1/med0$BUGSoutput$sims.list$r))*(1/apply(exp(mu_y1.2),2,expp, 1/med0$BUGSoutput$sims.list$r)))
tmean3=t(apply(vec,1,weib.trunc.mean,right=tmax))
if(multi)
tmean3=log(tmean3)
vec=cbind(med0$BUGSoutput$sims.list$r,
as.vector(med0$BUGSoutput$sims.list$lambda^(-1/med0$BUGSoutput$sims.list$r))*(1/apply(exp(mu_y0.2),2,expp, 1/med0$BUGSoutput$sims.list$r)))
tmean1=t(apply(vec,1,weib.trunc.mean,right=tmax))
if(multi)
tmean1=log(tmean1)
ie4[,,contm[j],l]=te4-(tmean3-tmean1)/deltap[l]
}
}}
#ie for binary mediators
if(p2>0){
for (k in 1:p2){
temp1.1=temp1
temp1.2=temp2
temp1.1[2:(nrow(M1)+1),binm1[k],]=matrix(M1[j1,binm1[k]],N,N1)
temp1.2[2:(nrow(M1)+1),binm1[k],]=matrix(M1[j1,binm1[k]],N,N1) #j2/j1
if (is.null(cova)){
mu_y0.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x1.temp1) + t(apply(temp1.1,3,matrix.prod))
mu_y1.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x3.temp1) + t(apply(temp1.2,3,matrix.prod))}
else{
mu_y0.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x1.temp1) + med0$BUGSoutput$sims.list$eta%*%t(cova)+t(apply(temp1.1,3,matrix.prod))
mu_y1.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x3.temp1) + med0$BUGSoutput$sims.list$eta%*%t(cova)+t(apply(temp1.2,3,matrix.prod))}
if(y.type==1)
ie4[,,binm[k],l]=te4-(mu_y1.2-mu_y0.2)/deltap[l]
else if(y.type==2)
ie4[,,binm[k],l]=te4-(exp(mu_y1.2)/(1+exp(mu_y1.2))-exp(mu_y0.2)/(1+exp(mu_y0.2)))/deltap[l]
else if (y.type==4)
{mu_y0.2<- mu_y0.2-matrix(med0$BUGSoutput$sims.list$beta0,N1,N)
mu_y1.2<- mu_y1.2-matrix(med0$BUGSoutput$sims.list$beta0,N1,N)
vec=cbind(med0$BUGSoutput$sims.list$r,
as.vector(med0$BUGSoutput$sims.list$lambda^(-1/med0$BUGSoutput$sims.list$r))*(1/apply(exp(mu_y1.2),2,expp, 1/med0$BUGSoutput$sims.list$r)))
tmean3=t(apply(vec,1,weib.trunc.mean,right=tmax))
if(multi)
tmean3=log(tmean3)
vec=cbind(med0$BUGSoutput$sims.list$r,
as.vector(med0$BUGSoutput$sims.list$lambda^(-1/med0$BUGSoutput$sims.list$r))*(1/apply(exp(mu_y0.2),2,expp, 1/med0$BUGSoutput$sims.list$r)))
tmean1=t(apply(vec,1,weib.trunc.mean,right=tmax))
if(multi)
tmean1=log(tmean1)
ie4[,,binm[k],l]=te4-(tmean3-tmean1)/deltap[l]
}
}}
#ie for categorical mediators
if(p3>0){
for (j in 1:p3){
temp1.1=temp1
temp1.2=temp2
for (i in catm1[j,1]:catm1[j,2])
{temp1.1[2:(nrow(M1)+1),i,]=matrix(M1[j1,i],N,N1)
temp1.2[2:(nrow(M1)+1),i,]=matrix(M1[j1,i],N,N1)} #j2/j1
if(is.null(cova)){
mu_y0.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x1.temp1) + t(apply(temp1.1,3,matrix.prod))
mu_y1.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x3.temp1) + t(apply(temp1.2,3,matrix.prod))}
else{
mu_y0.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x1.temp1) + med0$BUGSoutput$sims.list$eta%*%t(cova)+t(apply(temp1.1,3,matrix.prod))
mu_y1.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x3.temp1) + med0$BUGSoutput$sims.list$eta%*%t(cova)+t(apply(temp1.2,3,matrix.prod))}
if(y.type==1)
ie4[,,catm[j],l]=te4-(mu_y1.2-mu_y0.2)/deltap[l]
else if(y.type==2)
ie4[,,catm[j],l]=te4-(exp(mu_y1.2)/(1+exp(mu_y1.2))-exp(mu_y0.2)/(1+exp(mu_y0.2)))/deltap[l]
else if (y.type==4)
{mu_y0.2<- mu_y0.2-matrix(med0$BUGSoutput$sims.list$beta0,N1,N)
mu_y1.2<- mu_y1.2-matrix(med0$BUGSoutput$sims.list$beta0,N1,N)
vec=cbind(med0$BUGSoutput$sims.list$r,
as.vector(med0$BUGSoutput$sims.list$lambda^(-1/med0$BUGSoutput$sims.list$r))*(1/apply(exp(mu_y1.2),2,expp, 1/med0$BUGSoutput$sims.list$r)))
tmean3=t(apply(vec,1,weib.trunc.mean,right=tmax))
if(multi)
tmean3=log(tmean3)
vec=cbind(med0$BUGSoutput$sims.list$r,
as.vector(med0$BUGSoutput$sims.list$lambda^(-1/med0$BUGSoutput$sims.list$r))*(1/apply(exp(mu_y0.2),2,expp, 1/med0$BUGSoutput$sims.list$r)))
tmean1=t(apply(vec,1,weib.trunc.mean,right=tmax))
if(multi)
tmean1=log(tmean1)
ie4[,,catm[j],l]=te4-(tmean3-tmean1)/deltap[l]
}
}}
aie4[,,l]<-apply(array(ie4[,,,l],c(N1,N,p1+p2+p3)),c(1,3),mean,na.rm=T)
#4.3. Calculate the de
temp1.1=temp1
temp1.2=temp2
for (i in 1:ncol(M1))
{temp1.1[2:(nrow(M1)+1),i,]=matrix(M1[j1,i],N,N1)
temp1.2[2:(nrow(M1)+1),i,]=matrix(M1[j1,i],N,N1)} #j2/j1
if(is.null(cova)){
mu_y0.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x1.temp1) + t(apply(temp1.1,3,matrix.prod))
mu_y1.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x3.temp1) + t(apply(temp1.2,3,matrix.prod))}
else{
mu_y0.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x1.temp1) + med0$BUGSoutput$sims.list$eta%*%t(cova)+t(apply(temp1.1,3,matrix.prod))
mu_y1.2<- matrix(med0$BUGSoutput$sims.list$beta0,N1,N) + med0$BUGSoutput$sims.list$c%*%t(x3.temp1) + med0$BUGSoutput$sims.list$eta%*%t(cova)+t(apply(temp1.2,3,matrix.prod))}
if(y.type==1)
de4=(mu_y1.2-mu_y0.2)/deltap[l]
else if(y.type==2)
de4=(exp(mu_y1.2)/(1+exp(mu_y1.2))-exp(mu_y0.2)/(1+exp(mu_y0.2)))/deltap[l]
else if (y.type==4)
{mu_y0.2<- mu_y0.2-matrix(med0$BUGSoutput$sims.list$beta0,N1,N)
mu_y1.2<- mu_y1.2-matrix(med0$BUGSoutput$sims.list$beta0,N1,N)
vec=cbind(med0$BUGSoutput$sims.list$r,
as.vector(med0$BUGSoutput$sims.list$lambda^(-1/med0$BUGSoutput$sims.list$r))*(1/apply(exp(mu_y1.2),2,expp, 1/med0$BUGSoutput$sims.list$r)))
tmean3=t(apply(vec,1,weib.trunc.mean,right=tmax))
if(multi)
tmean3=log(tmean3)
vec=cbind(med0$BUGSoutput$sims.list$r,
as.vector(med0$BUGSoutput$sims.list$lambda^(-1/med0$BUGSoutput$sims.list$r))*(1/apply(exp(mu_y0.2),2,expp, 1/med0$BUGSoutput$sims.list$r)))
tmean1=t(apply(vec,1,weib.trunc.mean,right=tmax))
if(multi)
tmean1=log(tmean1)
de4=(tmean3-tmean1)/deltap[l]
}
ade4[,l]=apply(de4,1,mean,na.rm=T)
}
}
if(y.type==1)
de2=de1
ate1=apply(aie1,c(1,3),sum)+de1
ate2=apply(aie2,c(1,3),sum)+de2
colnames(aie4)=colnames(M2)
colnames(aie1)=colnames(M2)
colnames(aie2)=colnames(M2)
colnames(aie3)=colnames(M2)
}
#detach(med0$BUGSoutput$sims.list)
result=list(aie1=aie1, ade1=de1, ate1=ate1, aie2=aie2, ade2=de2, ate2=ate2,
aie3=aie3, ade3=de3, ate3=ate3, aie4=aie4, ade4=ade4, ate4=ate4,
sims.list=med0,data0=data0,omu3=omu3,omu4=omu4) #ie2=ie2, med0$BUGSoutput$sims.list
class(result)='bma.bx.cy'
return(result)
}
summary.bma.bx.cy<-function(object, ..., plot= TRUE, RE=TRUE, quant=c(0.025, 0.25, 0.5, 0.75,0.975),digit=4,method=3)
{
summary.med<-function(vec,qua=quant, digit=digit)
{c(mean=mean(vec,na.rm=T),sd=sd(vec,na.rm=T),quantile(vec,qua,na.rm=T))
}
summary.med.re<-function(vec,vec1,qua=quant, digit=digit)
{vec=vec/vec1
c(mean=mean(vec,na.rm=T),sd=sd(vec,na.rm=T),quantile(vec,qua,na.rm=T))}
if(object$data0$y_type!=3){
result1<-array(0,c(7,2+ncol(object$aie1),ncol(object$ade1)))
result1.re<-array(0,c(7,1+ncol(object$aie1),ncol(object$ade1)))
result2<-result1
result2.re<-result1.re
result3<-result1
result3.re<-result1.re
result4<-result1
result4.re<-result1.re
for(j in 1:ncol(object$ade1)){
result1[,,j]<-apply(cbind(TE=object$ate1[,j],DE=object$ade1[,j],object$aie1[,,j]),2,summary.med)
result1.re[,,j]<-apply(cbind(DE=object$ade1[,j],object$aie1[,,j]),2,summary.med.re,object$ate1[,j])
result2[,,j]<-apply(cbind(TE=object$ate2[,j],DE=object$ade2[,j],object$aie2[,,j]),2,summary.med)
result2.re[,,j]<-apply(cbind(DE=object$ade2[,j],object$aie2[,,j]),2,summary.med.re,object$ate2[,j])
result3[,,j]<-apply(cbind(TE=object$ate3[,j],DE=object$ade3[,j],object$aie3[,,j]),2,summary.med)
result3.re[,,j]<-apply(cbind(DE=object$ade3[,j],object$aie3[,,j]),2,summary.med.re,object$ate3[,j])
result4[,,j]<-apply(cbind(TE=object$ate4[,j],DE=object$ade4[,j],object$aie4[,,j]),2,summary.med)
result4.re[,,j]<-apply(cbind(DE=object$ade4[,j],object$aie4[,,j]),2,summary.med.re,object$ate4[,j])}}
else{
result1<-array(0,c(7,2+ncol(object$aie1),ncol(object$ade1),dim(object$ade1)[3]))
result1.re<-array(0,c(7,1+ncol(object$aie1),ncol(object$ade1),dim(object$ade1)[3]))
result2<-result1
result2.re<-result1.re
result3<-result1
result3.re<-result1.re
result4<-result1
result4.re<-result1.re
for(j in 1:ncol(object$ade1)){
for(q1 in 1:dim(object$ade1)[3]){
result1[,,j,q1]<-apply(cbind(TE=object$ate1[,j,q1],DE=object$ade1[,j,q1],object$aie1[,,j,q1]),2,summary.med)
result1.re[,,j,q1]<-apply(cbind(DE=object$ade1[,j,q1],object$aie1[,,j,q1]),2,summary.med.re,object$ate1[,j,q1])
result2[,,j,q1]<-apply(cbind(TE=object$ate2[,j,q1],DE=object$ade2[,j,q1],object$aie2[,,j,q1]),2,summary.med)
result2.re[,,j,q1]<-apply(cbind(DE=object$ade2[,j,q1],object$aie2[,,j,q1]),2,summary.med.re,object$ate2[,j,q1])
result3[,,j,q1]<-apply(cbind(TE=object$ate3[,j,q1],DE=object$ade3[,j,q1],object$aie3[,,j,q1]),2,summary.med)
result3.re[,,j,q1]<-apply(cbind(DE=object$ade3[,j,q1],object$aie3[,,j,q1]),2,summary.med.re,object$ate3[,j,q1])
result4[,,j,q1]<-apply(cbind(TE=object$ate4[,j,q1],DE=object$ade4[,j,q1],object$aie4[,,j,q1]),2,summary.med)
result4.re[,,j,q1]<-apply(cbind(DE=object$ade4[,j,q1],object$aie4[,,j,q1]),2,summary.med.re,object$ate4[,j,q1])}
}
}
c.names=colnames(object$aie1)
colnames(result1)=c("TE","DE",c.names)
colnames(result2)=c("TE","DE",c.names)
colnames(result3)=c("TE","DE",c.names)
colnames(result4)=c("TE","DE",c.names)
colnames(result1.re)=c("DE",c.names)
colnames(result2.re)=c("DE",c.names)
colnames(result3.re)=c("DE",c.names)
colnames(result4.re)=c("DE",c.names)
rownames(result1)=c("mean","sd",paste("q",quant,sep="_"))
rownames(result2)=c("mean","sd",paste("q",quant,sep="_"))
rownames(result3)=c("mean","sd",paste("q",quant,sep="_"))
rownames(result4)=c("mean","sd",paste("q",quant,sep="_"))
rownames(result1.re)=c("mean","sd",paste("q",quant,sep="_"))
rownames(result2.re)=c("mean","sd",paste("q",quant,sep="_"))
rownames(result3.re)=c("mean","sd",paste("q",quant,sep="_"))
rownames(result4.re)=c("mean","sd",paste("q",quant,sep="_"))
result=list(result1=result1,result1.re=result1.re,result2=result2,result2.re=result2.re,
result3=result3,result3.re=result3.re,result4=result4,result4.re=result4.re,method=method,
digit=digit,plot=plot,RE=RE,y.type=object$data0$y_type)
class(result)="summary.bma"
result
}
print.summary.bma<-function (x, ..., digit = x$digit, method=x$method, RE=x$RE)
{plot.sum<-function(obj1,main1="Estimated Effects")
{oldpar <- par(no.readonly = TRUE)
on.exit(par(oldpar))
re <- obj1[1,]
upper <- obj1[7,]
lower <- obj1[3,]
name1 <- colnames(obj1)
par(mfrow = c(1, 1), mar = c(1, 6, 1, 1), oma = c(3, 2, 2, 4))
bp <- barplot2(re, horiz = TRUE, main = main1,
names.arg = name1, plot.ci = TRUE, ci.u = upper,
ci.l = lower, cex.names = 0.9, beside = FALSE,
cex.axis = 0.9, las = 1, xlim = range(c(upper,lower), na.rm = TRUE,finite=T),
col = rainbow(length(re), start = 3/6, end = 4/6))}
if(x$y.type!=3){
for (j in 1:dim(x$result1)[3])
{cat('\n For Predictor', j, ':\n')
if (method==1)
{if(!RE)
{cat('Estimated Effects for Method 1:\n')
print(round(x$result1[,,j],digits = digit))
if(x$plot)
plot.sum(x$result1[,,j],main1=paste("Estimated Effects Using Method", method, "(predictor",j,")"))}
else{
cat('Estimated Relative Effects for Method 1:\n')
print(round(x$result1.re[,,j],digits = digit))
if(x$plot)
plot.sum(x$result1.re[,,j],main1=paste("Estimated Relative Effects Using Method", method, "(predictor",j,")"))}}
else if(method==2){
if(!RE)
{cat('Estimated Effects for Method 2:\n')
print(round(x$result2[,,j],digits = digit))
if(x$plot)
plot.sum(x$result2[,,j],main1=paste("Estimated Effects Using Method", method, "(predictor",j,")"))}
else{
cat('Estimated Relative Effects for Method 2:\n')
print(round(x$result2.re[,,j],digits = digit))
if(x$plot)
plot.sum(x$result2.re[,,j],main1=paste("Estimated Relative Effects Using Method", method, "(predictor",j,")"))}
}
else if(method==3){
if(!RE)
{cat('Estimated Effects for Method 3:\n')
print(round(x$result3[,,j],digits = digit))
if(x$plot)
plot.sum(x$result3[,,j],main1=paste("Estimated Effects Using Method", method, "(predictor",j,")"))}
else{
cat('Estimated Relative Effects for Method 3:\n')
print(round(x$result3.re[,,j],digits = digit))
if(x$plot)
plot.sum(x$result3.re[,,j],main1=paste("Estimated Relative Effects Using Method", method, "(predictor",j,")"))}
}
else if(method==4){
if(!RE)
{cat('Estimated Effects for Method 4:\n')
print(round(x$result4[,,j],digits = digit))
if(x$plot)
plot.sum(x$result4[,,j],main1=paste("Estimated Effects Using Method", method, "(predictor",j,")"))}
else{
cat('Estimated Relative Effects for Method 4:\n')
print(round(x$result4.re[,,j],digits = digit))
if(x$plot)
plot.sum(x$result4.re[,,j],main1=paste("Estimated Relative Effects Using Method", method, "(predictor",j,")"))}
}
}}
else{
for (j in 1:dim(x$result1)[3])
for (q1 in 1:dim(x$result1)[4])
{cat('\n For Predictor', j, 'outcome',q1, ':\n')
if (method==1)
{if(!RE)
{cat('Estimated Effects for Method 1:\n')
print(round(x$result1[,,j,q1],digits = digit))
if(x$plot)
plot.sum(x$result1[,,j,q1],main1=paste("Estimated Effects Using Method", method, "(predictor",j,"outcome",q1,")"))}
else{
cat('Estimated Relative Effects for Method 1:\n')
print(round(x$result1.re[,,j,q1],digits = digit))
if(x$plot)
plot.sum(x$result1.re[,,j,q1],main1=paste("Estimated Relative Effects Using Method", method, "(predictor",j,"outcome",q1,")"))}}
else if(method==2){
if(!RE)
{cat('Estimated Effects for Method 2:\n')
print(round(x$result2[,,j,q1],digits = digit))
if(x$plot)
plot.sum(x$result2[,,j,q1],main1=paste("Estimated Effects Using Method", method, "(predictor",j,"outcome",q1,")"))}
else{
cat('Estimated Relative Effects for Method 2:\n')
print(round(x$result2.re[,,j,q1],digits = digit))
if(x$plot)
plot.sum(x$result2.re[,,j,q1],main1=paste("Estimated Relative Effects Using Method", method, "(predictor",j,"outcome",q1,")"))}
}
else if(method==3){
if(!RE)
{cat('Estimated Effects for Method 3:\n')
print(round(x$result3[,,j,q1],digits = digit))
if(x$plot)
plot.sum(x$result3[,,j,q1],main1=paste("Estimated Effects Using Method", method, "(predictor",j,"outcome",q1,")"))}
else{
cat('Estimated Relative Effects for Method 3:\n')
print(round(x$result3.re[,,j,q1],digits = digit))
if(x$plot)
plot.sum(x$result3.re[,,j,q1],main1=paste("Estimated Relative Effects Using Method", method, "(predictor",j,"outcome",q1,")"))}
}
else if(method==4){
if(!RE)
{cat('Estimated Effects for Method 4:\n')
print(round(x$result4[,,j,q1],digits = digit))
if(x$plot)
plot.sum(x$result4[,,j,q1],main1=paste("Estimated Effects Using Method", method, "(predictor",j,"outcome",q1,")"))}
else{
cat('Estimated Relative Effects for Method 4:\n')
print(round(x$result4.re[,,j,q1],digits = digit))
if(x$plot)
plot.sum(x$result4.re[,,j,q1],main1=paste("Estimated Relative Effects Using Method", method, "(predictor",j,"outcome",q1,")"))}
}
}
}
}
Try the BayesianMediationA package in your browser
Any scripts or data that you put into this service are public.
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Embedding an R snippet on your website
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.