Nothing
R/ddpeffectsplot.R
In growcurves: Bayesian Semi and Nonparametric Growth Curve Models that Additionally Include Multiple Membership Random Effects
Defines functions
ddpEffectsplot
Documented in
ddpEffectsplot
#' Plot by-subject and by-treatment posterior mean values for dossage random effects
#'
#' Each \code{ddpgrow} object contains posterior mean estimates for the \code{q} x \code{T}
#' matrix of by-subject random effects that is extracted from the \code{ddpgrow}
#' object that is input to \code{ddpEffectsplot}.
#' This function produces a \code{q} x \code{T.m} heat map plot of posterior mean effect values
#' for the dosages in treatment \code{m} faceted on a set of chosen subjects. The resulting plot
#' produces a heatmap for each trt-subject combination. Both a ggplot2 plot object
#' and a \code{data.frame} object are returned.
#'
#' @param object A \code{ddpgrow} object.
#' @param subjects.plot A vector of subjects for performing plots that is composed of some subset of the \code{subject} vector input for modeling.
#' If left blank, a random subset is chosen from \code{subject}.
#' @param n.plot An optional scalar input for number of randomly generated subjects to plot (if \code{subjects.plot} is left blank).
#' @param trts.plot A vector of focus treatments to use for plotting.
#' @param x.axis.label An optional scalar character entry to label the treatment(s) dosages
#' @param smoother A scalar boolean input indicating whether to co-plot a smoother line with point values.
#' @param re.order A scalar boolean input indicating whether to sort the plots of effects in order of increasing value.
#' @param cred.intervals A boolean scalar indicating whether the by-subject effects plots should include credible intervals.
#' @param map.group A \code{matrix} or \code{data.frame} object containing a grouping of subjects that will be used
#' to produce an additional set of effect plots that aggregate subjects by the grouping structure.
#' The first column containing subject identifiers for all subjects modeled in \code{object}.
#' The second column contains the desired desired group identifiers that may be of type character or numeric.
#' @param n.dose.plot Optional numeric input for number of randomly chosen doses for which to plot effects growth curves.
#' @param orderto A numeric vector of length equal to the total number of dosages across all treatments that conveys an order to be used
#' by-dosage growth curve plots within cluster and treatment.
#' @return A list object containing a faceted set of heat maps (one per subject),
#' a faceted set of effect point plots, and the associated \code{data.frame} objct.
#' \item{dat.se}{A \code{data.frame} object used to generate the trt-subject faceted plots for effect means
#' Fields are titled, \code{c("order","dose","trt","subject","effects")}.}
#' \item{dat.ci}{A \code{data.frame} object used to generate the trt-subject faceted plots for effect credible intervals
#' Fields are titled, \code{c("order","dose","trt","subject","quantile","effects")}.}
#' \item{dat.clust}{A \code{data.frame} object used to generate the trt-group faceted plots for effect means
#' Fields are titled, \code{c("order","dose","trt","cluster","effects")}.}
#' \item{dat.clust.ci}{A \code{data.frame} object used to generate the trt-group faceted plots for effect credible intervals
#' Fields are titled, \code{c("order","dose","trt","cluster","quantile","effects")}.}
#' \item{dat.gc}{A \code{data.frame} object used to generate the by-dose growth curves (for multivariate polynomial effects)
#' Fields are titled, \code{c("fit","time","cluster","trt","dose")}.}
#' \item{p.hm}{A \code{ggplot2} object of heat maps for mean random effect values, faceted by trt and subject combinations.}
#' \item{p.pp}{A \code{ggplot2} object of point plots for mean random effect values or credible intervals, faceted by trt and subject combinations.}
#' \item{pc.m}{A \code{ggplot2} object of point plots for mean random effect values, faceted by trt and group combinations.}
#' \item{pc.ci}{A \code{ggplot2} object of point plots for credible intervals of random effect values, faceted by trt and cluster combinations.}
#' \item{pc.gc}{A \code{ggplot2} growth curve plots for each dose where the does effects are multivariate polynomial.}
#' @seealso \code{\link{ddpgrow}}, \code{\link{dpgrow}}, \code{\link{dpgrowmm}}, \code{\link{dpgrowmult}}
#' @author Terrance Savitsky \email{tds151@@gmail.com}
#' @aliases ddpEffectsplot
#' @export
ddpEffectsplot <- function(object, subjects.plot = NULL, n.plot = 3, trts.plot = NULL, x.axis.label = NULL, smoother = TRUE, re.order = TRUE, cred.intervals = TRUE, map.group = NULL, n.dose.plot = 5, orderto = NULL)
{
## check if object is of class ddpgrow
if( class(object) != "ddpgrow" ) stop("\nThe input 'object' must be of class 'ddpgrow' to use this function.\n")
## check if subjects.plot are in subjects modeled
map.subject = summary(object)$summary.results$map.subject ## data.frame
map.subject = unique(map.subject)
subjects.input = map.subject$label.input
if( is.null(subjects.plot) ) ## is.null(subjects.plot) = TRUE, randomly sample subject set
{
subjects.num = sample(1:nrow(map.subject), n.plot, replace = FALSE)
subjects.plot = subjects.num ## labels are same as numerical count of subjects used to model
}else{ ## check that subjects.plot is a strict subset of subjects.input
if( length(setdiff(subjects.plot,subjects.input)) > 0 ) stop("\nSubjects to plot must be a subset of those used in model.\n")
tmp = subset(map.subject, label.input %in% subjects.plot)
subjects.num = tmp$label.new ## user inputs names, we extract numerical equivalent used in modeling
rm(tmp)
}
## if( !is.null(map.group) ), check that subjects mapped are in subjects modeled
if( !is.null(map.group) )
{
map.group <- as.data.frame(map.group)
names(map.group) <- c("label.input","cluster")
if( length(setdiff(map.group$label.input,subjects.input)) > 0 ) stop("\nSubjects to group for plotting must be those used to model. Please check that first column of 'map.group' contains
subjects and second contains group membership.\n")
## replace label.input in map.group with numerical label.new
tmp <- merge(map.group, map.subject, by="label.input", all.x = TRUE, sort = FALSE)
map.newgroup <- data.frame(tmp$label.new,tmp$cluster)
names(map.newgroup) <- c("subjects.num","cluster")
## capture number of clusters
clusters <- unique(map.newgroup$cluster)
num.clust <- length(clusters)
}
## look for conflicts on re-ordering instructions
if( re.order == TRUE & !is.null(orderto) ) warning("\nSince re.order is set to TRUE, the orderto input will be ignored. If want to use orderto, then set re.order to FALSE.\n")
if( is.null(orderto) ) orderto = 1:sum(summary(object)$summary.results$numt) ## each entry in numt is the number of dosages associated to each treatment. length(numt) is the number of treatments.
## extract numerical trt as user inputs label
map.trtcov = summary(object)$summary.results$map.trtcov ## data.frame
map.trtcov = unique(map.trtcov)
if( !is.null(trts.plot) )
{
if( length(setdiff(map.trtcov$label.input,trts.plot)) > 0) stop("\nTreatments selected for plotting must be those used to model.\n")
tmp = subset(map.trtcov, label.input %in% trts.plot)
trt.num = tmp$label.new
rm(tmp)
}else{ ## is.null(trts.plot), so randomly select a set of treatments
typet = summary(object)$summary.results$typet
samp.treat = min(length(typet),2)
trt.num = sample(1:length(typet),samp.treat,replace=FALSE)
trts.plot = unique(map.trtcov$label.input)[trt.num]
}
## create plot data.frame
Nrandom = ncol(summary(object)$summary.results$Z)
Nsubject = nrow(unique(summary(object)$summary.results$map.subject))
numt = summary(object)$summary.results$numt
nt = length(trt.num)
np = length(subjects.num)
dat = vector("list",nt)
effect.mat = vector("list",nt)
for(m in 1:nt)
{
effect.mat[[m]] = summary(object)$summary.results$dosetrt.mean[[trt.num[m]]] ## select the trt.num[m] list element of an P x q*numt[m] matrix
effect.sum = summary(object)$summary.results$dosetrt.summary[[trt.num[m]]]
leff.m = t(matrix(effect.sum[,1], (Nrandom*numt[trt.num[m]]), Nsubject, byrow = FALSE))
heff.m = t(matrix(effect.sum[,3], (Nrandom*numt[trt.num[m]]), Nsubject, byrow = FALSE))
dose.m = colnames(effect.mat[[m]])[1:numt[trt.num[m]]]
dose.m = sapply(strsplit(dose.m,""),function(x){
x <- x[-c(1,2)] ## get rid of the "q." indicator of random effects in the names
x <- paste(x,collapse="")
x}) ## now we just have the names of the numt[trt.num[m]] doses associated to treatment trt.num[m]
dose.m = rep(dose.m, Nrandom)
order.m = rep(1:Nrandom, each = numt[trt.num[m]])
dat.m = vector("list",np)
for(i in 1:np)
{
effectsm.i = effect.mat[[m]][subjects.num[i],] ## extract subject (row), subjects.num[i], from the P x q*numt[m] effects.mat
leffm.i = leff.m[subjects.num[i],]
heffm.i = heff.m[subjects.num[i],]
dat.m[[i]] = data.frame(order.m, dose.m, trts.plot[m], subjects.plot[i], effectsm.i, leffm.i, heffm.i, stringsAsFactors = FALSE)
}
dat[[m]] = do.call("rbind",dat.m)
}
dat = do.call("rbind", dat)
names(dat) = c("order","dose","trt","subject","mean.effects","low.effects","high.effects")
## output plots
## tile geom
if( re.order == TRUE )
{
## first create variable, dose_trtsubject, that confines reordering to within trt-subject.
dat <- transform( dat, dose_trtsubject = factor(paste(dose,trt,subject,sep=".")) )
## although plotting dose_trtsubject, set x-axis tick labels to dose (within trt-subject)
options <- scale_x_discrete(labels=dat$dose, breaks=dat$dose_trtsubject)
p.t = ggplot(data=dat,aes(x=reorder(dose_trtsubject,mean.effects), y = factor(order), fill = mean.effects)) + options
}else{
p.t = ggplot(data=dat,aes(x=dose, y = factor(order), fill = mean.effects))
}
l = geom_tile()
yaxis = ylab("Polynomial Order")
if( is.null(x.axis.label) )
{
xaxis = xlab("Dose")
}else xaxis = xlab(eval(x.axis.label))
f = facet_wrap(trt~subject, scales="free", nrow=min(nt,5) )
dev.new()
p.t = p.t + l + xaxis + yaxis + f + labs(fill = "effects") + scale_fill_gradient(low = "white", high = "steelblue")
## + scale_fill_gradient( limits = c(-3002,-2980) )
print(p.t)
## point geom
if( cred.intervals == FALSE )
{
if( re.order == TRUE )
{
## first create variable, dose_trtsubject, that confines reordering to within trt-subject.
dat <- transform( dat, dose_trtsubject = factor(paste(dose,trt,subject,sep=".")) )
## although plotting dose_trtsubject, set x-axis tick labels to dose (within trt-subject)
options <- scale_x_discrete(labels=dat$dose, breaks=dat$dose_trtsubject)
p.p = ggplot(data=dat,aes(x=reorder(dose_trtsubject,mean.effects), y = mean.effects, colour = factor(order), shape = factor(order))) + options
}else{
p.p = ggplot(data=dat,aes(x=dose, y = mean.effects, colour = factor(order), shape = factor(order)))
}
l = geom_point()
l.2 = geom_smooth(aes(group=factor(order)), method = "loess", span = 1.0, alpha = 0.1, se = FALSE)##, colour = "pink")
yaxis = ylab("Effect Values")
options = labs(colour = "Order", shape="Order")
f = facet_wrap(trt~subject, scales="free", nrow=min(nt,n.plot) )
}else{
dat.ci = melt(dat, id = c("order","dose","trt","subject","dose_trtsubject"))
dat.ci$quantile = dat.ci$variable
dat.ci$effects = as.numeric(dat.ci$value)
dat.ci$variable <- dat.ci$value <- NULL
if( re.order == TRUE )
{
## first create variable, dose_trtsubject, that confines reordering to within trt-subject.
dat.ci <- transform( dat.ci, dose_trtsubject = factor(paste(dose,trt,subject,sep=".")) )
## although plotting dose_trtsubject, set x-axis tick labels to dose (within trt-subject)
options <- scale_x_discrete(labels=dat.ci$dose, breaks=dat.ci$dose_trtsubject)
p.p = ggplot(data=dat.ci,aes(x=reorder(dose_trtsubject,effects), y = effects, colour = factor(order), shape = factor(order))) + options
}else{
p.p = ggplot(data=dat.ci,aes(x=dose, y = effects, colour = factor(order), shape = factor(order)))
}
l = geom_point(na.rm=TRUE)
## l = geom_line
l.2 = geom_smooth(aes(group=factor(order)), method = "loess", span = 1.0, alpha = 0.1,
se = FALSE, na.rm=TRUE)##, colour = "pink")
yaxis = ylab("Effect Values")
options = labs(colour = "Order", shape="Order")
f = facet_wrap(trt~subject, scales="free", nrow=min(nt,n.plot) )
}
if( smoother == TRUE)
{
p.p = p.p + l + l.2 + xaxis + yaxis + f + options + theme(axis.text.x=element_text(size=6, angle = 90))
}else{
p.p = p.p + l + xaxis + yaxis + f + options + theme(axis.text.x=element_text(size=6, angle = 90))
}
dev.new()
print(p.p)
if(!is.null(map.group)) ## plot subject effects aggregated to group
{
dat.clust = vector("list",nt)
dose.lab = vector("list",nt)
dose.ord = vector("list",nt)
md <- sapply(1:nt,function(m){md = numt[trt.num[m]]}) ## vector of dosages for each plotted treatment.
start.m <- 1 ## to produce user-input ordering of dosages for dosage growth curve plots (within cluster and treatment).
for( m in 1:nt )
{
## develop posterior draws for cluster effects means
DoseEffects.m <- object$DoseEffects[[trt.num[m]]] ## B x Nsubject*(Nrandom*numt[m])
tDE.m <- t(DoseEffects.m)
## some dimensions
step.m <- Nrandom*numt[trt.num[m]] ## effects per subject (and cluster once we average over subjects_
B <- nrow(DoseEffects.m)
## some structures
tDC.m <- matrix(0,(num.clust*step.m),B)
for( c in 1:num.clust )
{
subj.c <- subset(map.newgroup, cluster == clusters[c])$subjects.num
nums.c <- length(subj.c)
cols.c <- vector("numeric",(nums.c*step.m))
for( s in 1:nums.c )
{
indx.s <- ((s-1)*step.m + 1):(s*step.m)
cols.c[indx.s] <- c( ((subj.c[s]-1)*step.m + 1):(subj.c[s]*step.m) )
}
indx.c <- ((c-1)*step.m + 1):(c*step.m)
## we use 'rowsum' to chunk sum for each member in step.m and then transpose the result
chunk.c <- rep(1:step.m, times = nums.c) ## recall subjects are the slow moving index
## composing average of the step.m block of effects by subjects in cluster c for each posterior sample in B
tDC.m[indx.c,] <- rowsum(tDE.m[cols.c,],group = chunk.c) * (1/nums.c) ## num.clust*step.m x B
}
DoseCluster.m <- t(tDC.m) ## B x num.clust*step.m
## compose quantile summaries
mcmc.data = DoseCluster.m
mcmc.mean = colMeans(mcmc.data)
mcmc.low = apply(mcmc.data,2,function(x){quantile(x,probs = 0.025)})
mcmc.high = apply(mcmc.data,2,function(x){quantile(x,probs = 0.975)})
## for doseclust.summary.m, cluster is slowest moving index, followed by polynomial order, followed by numt[m]
## <--> num.clust*(Nrandom*numt[m]) x 1, index speed order: cluster, order, dose
## compose dose effect quantiles for the clusters - has num.clust rows (which is probably pretty small)
doseclust.summary.m = t(rbind(mcmc.low,mcmc.mean,mcmc.high)) ## transpose to 3 columns for c("mcmc.low","mcmc.mean","mcmc.high")
## define indices
order.m = rep(1:Nrandom, each = numt[trt.num[m]], times = num.clust) ## of polynomial order
dose.m = colnames(effect.mat[[m]])[1:numt[trt.num[m]]] ## use effect.mat from above b/c contains dose names
dose.mlab = sapply(strsplit(dose.m,""),function(x){
x <- x[-c(1,2)] ## get rid of the "q." indicator of random effects in the names
x <- paste(x,collapse="")
x}) ## now we just have the names of the numt[trt.num[m]] doses associated to treatment trt.num[m]
end.m <- start.m + md[m] - 1
dose.mord = orderto[start.m:end.m] ## ordering dose names across treatments
start.m <- end.m + 1
dose.m = rep(dose.mlab, times = (Nrandom*num.clust))
clust.m = rep(clusters,each=(Nrandom*numt[trt.num[m]]))
dat.clust[[m]] <- data.frame(order.m, dose.m, trts.plot[m], clust.m, doseclust.summary.m, stringsAsFactors = FALSE)
dose.lab[[m]] <- dose.mlab
dose.ord[[m]] <- dose.mord
} ## end loop over nt treatments
dat.clust = do.call("rbind", dat.clust)
names(dat.clust) = c("order","dose","trt","cluster","low.effects","mean.effects","high.effects")
## perform point plotting
## of means
if( re.order == TRUE )
{
## first create variable, dose_trtcluster, that confines reordering to within trt-cluster.
dat.clust <- transform( dat.clust, dose_trtclust = factor(paste(dose,trt,cluster,sep=".")) )
## although plotting dose_trtclust, set x-axis tick labels to dose (within trt-cluster)
options <- scale_x_discrete(labels=dat.clust$dose, breaks=dat.clust$dose_trtclust)
pc.m = ggplot(data=dat.clust,aes(x=reorder(dose_trtclust,mean.effects), y = mean.effects, colour = factor(order), shape = factor(order))) + options
}else{
pc.m = ggplot(data=dat.clust,aes(x = dose, y = mean.effects, colour = factor(order), shape = factor(order)))
}
l = geom_point()
l.2 = geom_smooth(aes(linetype = factor(order), group=factor(order)), method = "loess", span = 1.0, alpha = 0.1, se = FALSE)##, colour = "pink")
yaxis = ylab("Effect Values")
options = labs(colour = "Order", shape="Order",linetype = "Order")
f = facet_wrap( trt~cluster, scales = "free_x", nrow=nt )
if( smoother == TRUE)
{
pc.m = pc.m + l + l.2 + xaxis + yaxis + f + options + theme_bw() + theme(axis.text.x=element_text(size=6, angle = 90))
}else{
pc.m = pc.m + l + xaxis + yaxis + f + options + theme_bw() + theme(axis.text.x=element_text(size=6, angle = 90))
}
dev.new()
print(pc.m)
if( re.order == TRUE )
{
## of credible intervals, by cluster
datclust.ci = melt(dat.clust, id = c("order","dose","trt","cluster","dose_trtclust"))
datclust.ci$quantile = datclust.ci$variable
datclust.ci$effects = datclust.ci$value
datclust.ci$variable <- datclust.ci$value <- NULL
## although plotting dose_trtclust, set x-axis tick labels to dose (within trt-cluster)
options <- scale_x_discrete(labels=datclust.ci$dose, breaks=datclust.ci$dose_trtclust)
pc.ci = ggplot(data=datclust.ci,aes(x=reorder(dose_trtclust,effects), y = effects, colour = factor(order), shape = factor(order))) + options
}else{
## of credible intervals, by cluster
datclust.ci = melt(dat.clust, id = c("order","dose","trt","cluster"))
datclust.ci$quantile = datclust.ci$variable
datclust.ci$effects = datclust.ci$value
datclust.ci$variable <- datclust.ci$value <- NULL
pc.ci = ggplot(data=datclust.ci,aes(x = dose, y = effects, colour = factor(order), shape = factor(order)))
}
l = geom_line()
l.2 = geom_smooth(aes(group=factor(order)), method = "loess", span = 1.0, alpha = 0.1, se = FALSE)##, colour = "pink")
yaxis = ylab("Effects")
options.2 = labs(colour = "Order", shape="Order")
f = facet_wrap( trt~cluster, scales= "free_x", nrow=nt )
if( smoother == TRUE)
{
pc.ci = pc.ci + l + l.2 + xaxis + yaxis + f + options.2 + theme(axis.text.x=element_text(size=6, angle = 90))
}else{
pc.ci = pc.ci + l + xaxis + yaxis + f + options.2 + theme(axis.text.x=element_text(size=6, angle = 90))
}
dev.new()
print(pc.ci)
if( Nrandom > 1 ) ## more than one effect per dose
{
##
## plot within sample predicted growth curves for each dose, within cluster
##
## set-up structures
C <- length(clusters)
M <- length(trts.plot)
MD <- sum( md )
md.cum <- cumsum( md )
T <- 10 ## length of time intervals
cluster.gc <- trt.gc <- dose.gc <- vector("character", C*MD*T)
y.gc <- time.gc <- dose.order <- vector("numeric", C*MD*T)
X <- object$summary.results$X
measure.times <- unique( X[,(colnames(X) == "time^1")] )
min.T <- min(measure.times); max.T <- max(measure.times)
time <- seq(min.T,max.T,length.out = T)
## compute "response" from effect orders for each dose within treatment within cluster at T timempoints
for( c in 1:C )
{
for( m in 1:M )
{
for( d in 1:md[m] )
{
dat.cmd <- subset( dat.clust, (cluster %in% clusters[c]) & (trt %in% trts.plot[m]) & (dose %in% dose.lab[[m]][d]) )
effect.cmd <- vector("numeric", Nrandom)
for( r in 1:Nrandom )
{
effect.cmd[r] <- dat.cmd$mean.effects[dat.cmd$order == r]
}
for( t in 1:T )
{
## fill return objects
indx.cmdt <- (c-1)*(MD*T) + ((md.cum[m]-md[m])*T) + (d-1)*T + t
cluster.gc[indx.cmdt] <- as.character(clusters[c])
dose.gc[indx.cmdt] <- dose.lab[[m]][d]
dose.order[indx.cmdt] <- dose.ord[[m]][d]
trt.gc[indx.cmdt] <- trts.plot[m]
time.gc[indx.cmdt] <- time[t]
## generate response
y.cmdt <- 0
for( r in 1:Nrandom )
{
y.cmdt <- y.cmdt + effect.cmd[r]*(time[t]^(r-1))
}
y.gc[indx.cmdt] <- y.cmdt
} ## end loop over T time points
} ## end loop over D doses
} ## end loop over M treatments
} ## end loop over C clusters
dat.gc = data.frame(y.gc,time.gc,cluster.gc,trt.gc,dose.gc,dose.order, stringsAsFactors = FALSE)
names(dat.gc) = c("fit","time","cluster","trt","dose","dose.order")
## select doses to plot
dose.plot <- vector("list",nt)
for(m in 1:nt) {dose.plot[[m]] <- sample(dose.lab[[m]],min(length(dose.lab[[m]]),n.dose.plot),replace = FALSE)}
dose.plot <- unlist(dose.plot)
## dose.plot <- c(4,5,6,7,9,10,11,12,16,18,22,23,25,26,29,35,42,45,47,52)
dat.gcplot <- subset(dat.gc, dose %in% dose.plot)
n.clust <- length(unique(dat.gcplot$cluster))
##
## for ALL clients - MEAN growth curve, by TREATMENT category
##
p.gc = ggplot(data=dat.gcplot,aes(x=time, y = fit, colour = factor(reorder(dose,dose.order)))) ## here, looking for individual subject visibility
## l = geom_line(aes(linetype = factor(dose)))
l = geom_line()
l.2 = geom_point()
## l.2 = geom_smooth(aes(group=1),method = "loess", size = 1.1, colour = "black")
axis = labs(x = "Time", y = "Response Units", colour = eval(x.axis.label), linetype = eval(x.axis.label), shape = eval(x.axis.label)) ## x.axis.label captures "dose" label
f = facet_wrap(trt~cluster, ncol = n.clust, scales="fixed")
p.gc = p.gc + l + l.2 + f + axis + scale_colour_grey(start = 0.1, end = 0.5) + theme_bw()
dev.new()
print(p.gc)
}else{ ## Nrandom = 1, so don't plot a growth curve
p.gc <- NULL
dat.gc <- NULL
} ## end conditional statement on Nrandom > 1 to plot by-dose effect growth curves
} ## end conditional statement on whether is.null(map.group)
## return objects
if( is.null(map.group) )
{
if( cred.intervals == FALSE )
{
return(invisible(list(dat.se = dat, p.hm = p.t, p.pp = p.p)))
}else{
return(invisible(list(dat.se = dat, dat.ci = dat.ci, p.hm = p.t, p.pp = p.p)))
}
}else{
if( cred.intervals == FALSE )
{
return(invisible(list(dat.se = dat, dat.clust = dat.clust, dat.clust.ci = datclust.ci, dat.gc = dat.gc, p.hm = p.t, p.pp = p.p, pc.m = pc.m, pc.ci = pc.ci, p.gc = p.gc)))
}else{
return(invisible(list(dat.se = dat, dat.ci = dat.ci, dat.clust = dat.clust, dat.clust.ci = datclust.ci, dat.gc = dat.gc, p.hm = p.t, p.pp = p.p, pc.m = pc.m, pc.ci = pc.ci, p.gc = p.gc)))
}
}
subject <- subjects.num <- label.input <- mean.effects <- dose <- cluster <- fit <- trt <- dose_trtsubject <- dose_trtclust <- effects <- quantile <- NULL
} ## end function ddpEffectsplot
Try the growcurves package in your browser
Any scripts or data that you put into this service are public.
growcurves documentation built on May 2, 2019, 7:03 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 ddpEffectsplot
Documented in ddpEffectsplot
#' Plot by-subject and by-treatment posterior mean values for dossage random effects
#'
#' Each \code{ddpgrow} object contains posterior mean estimates for the \code{q} x \code{T}
#' matrix of by-subject random effects that is extracted from the \code{ddpgrow}
#' object that is input to \code{ddpEffectsplot}.
#' This function produces a \code{q} x \code{T.m} heat map plot of posterior mean effect values
#' for the dosages in treatment \code{m} faceted on a set of chosen subjects. The resulting plot
#' produces a heatmap for each trt-subject combination. Both a ggplot2 plot object
#' and a \code{data.frame} object are returned.
#'
#' @param object A \code{ddpgrow} object.
#' @param subjects.plot A vector of subjects for performing plots that is composed of some subset of the \code{subject} vector input for modeling.
#' If left blank, a random subset is chosen from \code{subject}.
#' @param n.plot An optional scalar input for number of randomly generated subjects to plot (if \code{subjects.plot} is left blank).
#' @param trts.plot A vector of focus treatments to use for plotting.
#' @param x.axis.label An optional scalar character entry to label the treatment(s) dosages
#' @param smoother A scalar boolean input indicating whether to co-plot a smoother line with point values.
#' @param re.order A scalar boolean input indicating whether to sort the plots of effects in order of increasing value.
#' @param cred.intervals A boolean scalar indicating whether the by-subject effects plots should include credible intervals.
#' @param map.group A \code{matrix} or \code{data.frame} object containing a grouping of subjects that will be used
#' to produce an additional set of effect plots that aggregate subjects by the grouping structure.
#' The first column containing subject identifiers for all subjects modeled in \code{object}.
#' The second column contains the desired desired group identifiers that may be of type character or numeric.
#' @param n.dose.plot Optional numeric input for number of randomly chosen doses for which to plot effects growth curves.
#' @param orderto A numeric vector of length equal to the total number of dosages across all treatments that conveys an order to be used
#' by-dosage growth curve plots within cluster and treatment.
#' @return A list object containing a faceted set of heat maps (one per subject),
#' a faceted set of effect point plots, and the associated \code{data.frame} objct.
#' \item{dat.se}{A \code{data.frame} object used to generate the trt-subject faceted plots for effect means
#' Fields are titled, \code{c("order","dose","trt","subject","effects")}.}
#' \item{dat.ci}{A \code{data.frame} object used to generate the trt-subject faceted plots for effect credible intervals
#' Fields are titled, \code{c("order","dose","trt","subject","quantile","effects")}.}
#' \item{dat.clust}{A \code{data.frame} object used to generate the trt-group faceted plots for effect means
#' Fields are titled, \code{c("order","dose","trt","cluster","effects")}.}
#' \item{dat.clust.ci}{A \code{data.frame} object used to generate the trt-group faceted plots for effect credible intervals
#' Fields are titled, \code{c("order","dose","trt","cluster","quantile","effects")}.}
#' \item{dat.gc}{A \code{data.frame} object used to generate the by-dose growth curves (for multivariate polynomial effects)
#' Fields are titled, \code{c("fit","time","cluster","trt","dose")}.}
#' \item{p.hm}{A \code{ggplot2} object of heat maps for mean random effect values, faceted by trt and subject combinations.}
#' \item{p.pp}{A \code{ggplot2} object of point plots for mean random effect values or credible intervals, faceted by trt and subject combinations.}
#' \item{pc.m}{A \code{ggplot2} object of point plots for mean random effect values, faceted by trt and group combinations.}
#' \item{pc.ci}{A \code{ggplot2} object of point plots for credible intervals of random effect values, faceted by trt and cluster combinations.}
#' \item{pc.gc}{A \code{ggplot2} growth curve plots for each dose where the does effects are multivariate polynomial.}
#' @seealso \code{\link{ddpgrow}}, \code{\link{dpgrow}}, \code{\link{dpgrowmm}}, \code{\link{dpgrowmult}}
#' @author Terrance Savitsky \email{tds151@@gmail.com}
#' @aliases ddpEffectsplot
#' @export
ddpEffectsplot <- function(object, subjects.plot = NULL, n.plot = 3, trts.plot = NULL, x.axis.label = NULL, smoother = TRUE, re.order = TRUE, cred.intervals = TRUE, map.group = NULL, n.dose.plot = 5, orderto = NULL)
{
## check if object is of class ddpgrow
if( class(object) != "ddpgrow" ) stop("\nThe input 'object' must be of class 'ddpgrow' to use this function.\n")
## check if subjects.plot are in subjects modeled
map.subject = summary(object)$summary.results$map.subject ## data.frame
map.subject = unique(map.subject)
subjects.input = map.subject$label.input
if( is.null(subjects.plot) ) ## is.null(subjects.plot) = TRUE, randomly sample subject set
{
subjects.num = sample(1:nrow(map.subject), n.plot, replace = FALSE)
subjects.plot = subjects.num ## labels are same as numerical count of subjects used to model
}else{ ## check that subjects.plot is a strict subset of subjects.input
if( length(setdiff(subjects.plot,subjects.input)) > 0 ) stop("\nSubjects to plot must be a subset of those used in model.\n")
tmp = subset(map.subject, label.input %in% subjects.plot)
subjects.num = tmp$label.new ## user inputs names, we extract numerical equivalent used in modeling
rm(tmp)
}
## if( !is.null(map.group) ), check that subjects mapped are in subjects modeled
if( !is.null(map.group) )
{
map.group <- as.data.frame(map.group)
names(map.group) <- c("label.input","cluster")
if( length(setdiff(map.group$label.input,subjects.input)) > 0 ) stop("\nSubjects to group for plotting must be those used to model. Please check that first column of 'map.group' contains
subjects and second contains group membership.\n")
## replace label.input in map.group with numerical label.new
tmp <- merge(map.group, map.subject, by="label.input", all.x = TRUE, sort = FALSE)
map.newgroup <- data.frame(tmp$label.new,tmp$cluster)
names(map.newgroup) <- c("subjects.num","cluster")
## capture number of clusters
clusters <- unique(map.newgroup$cluster)
num.clust <- length(clusters)
}
## look for conflicts on re-ordering instructions
if( re.order == TRUE & !is.null(orderto) ) warning("\nSince re.order is set to TRUE, the orderto input will be ignored. If want to use orderto, then set re.order to FALSE.\n")
if( is.null(orderto) ) orderto = 1:sum(summary(object)$summary.results$numt) ## each entry in numt is the number of dosages associated to each treatment. length(numt) is the number of treatments.
## extract numerical trt as user inputs label
map.trtcov = summary(object)$summary.results$map.trtcov ## data.frame
map.trtcov = unique(map.trtcov)
if( !is.null(trts.plot) )
{
if( length(setdiff(map.trtcov$label.input,trts.plot)) > 0) stop("\nTreatments selected for plotting must be those used to model.\n")
tmp = subset(map.trtcov, label.input %in% trts.plot)
trt.num = tmp$label.new
rm(tmp)
}else{ ## is.null(trts.plot), so randomly select a set of treatments
typet = summary(object)$summary.results$typet
samp.treat = min(length(typet),2)
trt.num = sample(1:length(typet),samp.treat,replace=FALSE)
trts.plot = unique(map.trtcov$label.input)[trt.num]
}
## create plot data.frame
Nrandom = ncol(summary(object)$summary.results$Z)
Nsubject = nrow(unique(summary(object)$summary.results$map.subject))
numt = summary(object)$summary.results$numt
nt = length(trt.num)
np = length(subjects.num)
dat = vector("list",nt)
effect.mat = vector("list",nt)
for(m in 1:nt)
{
effect.mat[[m]] = summary(object)$summary.results$dosetrt.mean[[trt.num[m]]] ## select the trt.num[m] list element of an P x q*numt[m] matrix
effect.sum = summary(object)$summary.results$dosetrt.summary[[trt.num[m]]]
leff.m = t(matrix(effect.sum[,1], (Nrandom*numt[trt.num[m]]), Nsubject, byrow = FALSE))
heff.m = t(matrix(effect.sum[,3], (Nrandom*numt[trt.num[m]]), Nsubject, byrow = FALSE))
dose.m = colnames(effect.mat[[m]])[1:numt[trt.num[m]]]
dose.m = sapply(strsplit(dose.m,""),function(x){
x <- x[-c(1,2)] ## get rid of the "q." indicator of random effects in the names
x <- paste(x,collapse="")
x}) ## now we just have the names of the numt[trt.num[m]] doses associated to treatment trt.num[m]
dose.m = rep(dose.m, Nrandom)
order.m = rep(1:Nrandom, each = numt[trt.num[m]])
dat.m = vector("list",np)
for(i in 1:np)
{
effectsm.i = effect.mat[[m]][subjects.num[i],] ## extract subject (row), subjects.num[i], from the P x q*numt[m] effects.mat
leffm.i = leff.m[subjects.num[i],]
heffm.i = heff.m[subjects.num[i],]
dat.m[[i]] = data.frame(order.m, dose.m, trts.plot[m], subjects.plot[i], effectsm.i, leffm.i, heffm.i, stringsAsFactors = FALSE)
}
dat[[m]] = do.call("rbind",dat.m)
}
dat = do.call("rbind", dat)
names(dat) = c("order","dose","trt","subject","mean.effects","low.effects","high.effects")
## output plots
## tile geom
if( re.order == TRUE )
{
## first create variable, dose_trtsubject, that confines reordering to within trt-subject.
dat <- transform( dat, dose_trtsubject = factor(paste(dose,trt,subject,sep=".")) )
## although plotting dose_trtsubject, set x-axis tick labels to dose (within trt-subject)
options <- scale_x_discrete(labels=dat$dose, breaks=dat$dose_trtsubject)
p.t = ggplot(data=dat,aes(x=reorder(dose_trtsubject,mean.effects), y = factor(order), fill = mean.effects)) + options
}else{
p.t = ggplot(data=dat,aes(x=dose, y = factor(order), fill = mean.effects))
}
l = geom_tile()
yaxis = ylab("Polynomial Order")
if( is.null(x.axis.label) )
{
xaxis = xlab("Dose")
}else xaxis = xlab(eval(x.axis.label))
f = facet_wrap(trt~subject, scales="free", nrow=min(nt,5) )
dev.new()
p.t = p.t + l + xaxis + yaxis + f + labs(fill = "effects") + scale_fill_gradient(low = "white", high = "steelblue")
## + scale_fill_gradient( limits = c(-3002,-2980) )
print(p.t)
## point geom
if( cred.intervals == FALSE )
{
if( re.order == TRUE )
{
## first create variable, dose_trtsubject, that confines reordering to within trt-subject.
dat <- transform( dat, dose_trtsubject = factor(paste(dose,trt,subject,sep=".")) )
## although plotting dose_trtsubject, set x-axis tick labels to dose (within trt-subject)
options <- scale_x_discrete(labels=dat$dose, breaks=dat$dose_trtsubject)
p.p = ggplot(data=dat,aes(x=reorder(dose_trtsubject,mean.effects), y = mean.effects, colour = factor(order), shape = factor(order))) + options
}else{
p.p = ggplot(data=dat,aes(x=dose, y = mean.effects, colour = factor(order), shape = factor(order)))
}
l = geom_point()
l.2 = geom_smooth(aes(group=factor(order)), method = "loess", span = 1.0, alpha = 0.1, se = FALSE)##, colour = "pink")
yaxis = ylab("Effect Values")
options = labs(colour = "Order", shape="Order")
f = facet_wrap(trt~subject, scales="free", nrow=min(nt,n.plot) )
}else{
dat.ci = melt(dat, id = c("order","dose","trt","subject","dose_trtsubject"))
dat.ci$quantile = dat.ci$variable
dat.ci$effects = as.numeric(dat.ci$value)
dat.ci$variable <- dat.ci$value <- NULL
if( re.order == TRUE )
{
## first create variable, dose_trtsubject, that confines reordering to within trt-subject.
dat.ci <- transform( dat.ci, dose_trtsubject = factor(paste(dose,trt,subject,sep=".")) )
## although plotting dose_trtsubject, set x-axis tick labels to dose (within trt-subject)
options <- scale_x_discrete(labels=dat.ci$dose, breaks=dat.ci$dose_trtsubject)
p.p = ggplot(data=dat.ci,aes(x=reorder(dose_trtsubject,effects), y = effects, colour = factor(order), shape = factor(order))) + options
}else{
p.p = ggplot(data=dat.ci,aes(x=dose, y = effects, colour = factor(order), shape = factor(order)))
}
l = geom_point(na.rm=TRUE)
## l = geom_line
l.2 = geom_smooth(aes(group=factor(order)), method = "loess", span = 1.0, alpha = 0.1,
se = FALSE, na.rm=TRUE)##, colour = "pink")
yaxis = ylab("Effect Values")
options = labs(colour = "Order", shape="Order")
f = facet_wrap(trt~subject, scales="free", nrow=min(nt,n.plot) )
}
if( smoother == TRUE)
{
p.p = p.p + l + l.2 + xaxis + yaxis + f + options + theme(axis.text.x=element_text(size=6, angle = 90))
}else{
p.p = p.p + l + xaxis + yaxis + f + options + theme(axis.text.x=element_text(size=6, angle = 90))
}
dev.new()
print(p.p)
if(!is.null(map.group)) ## plot subject effects aggregated to group
{
dat.clust = vector("list",nt)
dose.lab = vector("list",nt)
dose.ord = vector("list",nt)
md <- sapply(1:nt,function(m){md = numt[trt.num[m]]}) ## vector of dosages for each plotted treatment.
start.m <- 1 ## to produce user-input ordering of dosages for dosage growth curve plots (within cluster and treatment).
for( m in 1:nt )
{
## develop posterior draws for cluster effects means
DoseEffects.m <- object$DoseEffects[[trt.num[m]]] ## B x Nsubject*(Nrandom*numt[m])
tDE.m <- t(DoseEffects.m)
## some dimensions
step.m <- Nrandom*numt[trt.num[m]] ## effects per subject (and cluster once we average over subjects_
B <- nrow(DoseEffects.m)
## some structures
tDC.m <- matrix(0,(num.clust*step.m),B)
for( c in 1:num.clust )
{
subj.c <- subset(map.newgroup, cluster == clusters[c])$subjects.num
nums.c <- length(subj.c)
cols.c <- vector("numeric",(nums.c*step.m))
for( s in 1:nums.c )
{
indx.s <- ((s-1)*step.m + 1):(s*step.m)
cols.c[indx.s] <- c( ((subj.c[s]-1)*step.m + 1):(subj.c[s]*step.m) )
}
indx.c <- ((c-1)*step.m + 1):(c*step.m)
## we use 'rowsum' to chunk sum for each member in step.m and then transpose the result
chunk.c <- rep(1:step.m, times = nums.c) ## recall subjects are the slow moving index
## composing average of the step.m block of effects by subjects in cluster c for each posterior sample in B
tDC.m[indx.c,] <- rowsum(tDE.m[cols.c,],group = chunk.c) * (1/nums.c) ## num.clust*step.m x B
}
DoseCluster.m <- t(tDC.m) ## B x num.clust*step.m
## compose quantile summaries
mcmc.data = DoseCluster.m
mcmc.mean = colMeans(mcmc.data)
mcmc.low = apply(mcmc.data,2,function(x){quantile(x,probs = 0.025)})
mcmc.high = apply(mcmc.data,2,function(x){quantile(x,probs = 0.975)})
## for doseclust.summary.m, cluster is slowest moving index, followed by polynomial order, followed by numt[m]
## <--> num.clust*(Nrandom*numt[m]) x 1, index speed order: cluster, order, dose
## compose dose effect quantiles for the clusters - has num.clust rows (which is probably pretty small)
doseclust.summary.m = t(rbind(mcmc.low,mcmc.mean,mcmc.high)) ## transpose to 3 columns for c("mcmc.low","mcmc.mean","mcmc.high")
## define indices
order.m = rep(1:Nrandom, each = numt[trt.num[m]], times = num.clust) ## of polynomial order
dose.m = colnames(effect.mat[[m]])[1:numt[trt.num[m]]] ## use effect.mat from above b/c contains dose names
dose.mlab = sapply(strsplit(dose.m,""),function(x){
x <- x[-c(1,2)] ## get rid of the "q." indicator of random effects in the names
x <- paste(x,collapse="")
x}) ## now we just have the names of the numt[trt.num[m]] doses associated to treatment trt.num[m]
end.m <- start.m + md[m] - 1
dose.mord = orderto[start.m:end.m] ## ordering dose names across treatments
start.m <- end.m + 1
dose.m = rep(dose.mlab, times = (Nrandom*num.clust))
clust.m = rep(clusters,each=(Nrandom*numt[trt.num[m]]))
dat.clust[[m]] <- data.frame(order.m, dose.m, trts.plot[m], clust.m, doseclust.summary.m, stringsAsFactors = FALSE)
dose.lab[[m]] <- dose.mlab
dose.ord[[m]] <- dose.mord
} ## end loop over nt treatments
dat.clust = do.call("rbind", dat.clust)
names(dat.clust) = c("order","dose","trt","cluster","low.effects","mean.effects","high.effects")
## perform point plotting
## of means
if( re.order == TRUE )
{
## first create variable, dose_trtcluster, that confines reordering to within trt-cluster.
dat.clust <- transform( dat.clust, dose_trtclust = factor(paste(dose,trt,cluster,sep=".")) )
## although plotting dose_trtclust, set x-axis tick labels to dose (within trt-cluster)
options <- scale_x_discrete(labels=dat.clust$dose, breaks=dat.clust$dose_trtclust)
pc.m = ggplot(data=dat.clust,aes(x=reorder(dose_trtclust,mean.effects), y = mean.effects, colour = factor(order), shape = factor(order))) + options
}else{
pc.m = ggplot(data=dat.clust,aes(x = dose, y = mean.effects, colour = factor(order), shape = factor(order)))
}
l = geom_point()
l.2 = geom_smooth(aes(linetype = factor(order), group=factor(order)), method = "loess", span = 1.0, alpha = 0.1, se = FALSE)##, colour = "pink")
yaxis = ylab("Effect Values")
options = labs(colour = "Order", shape="Order",linetype = "Order")
f = facet_wrap( trt~cluster, scales = "free_x", nrow=nt )
if( smoother == TRUE)
{
pc.m = pc.m + l + l.2 + xaxis + yaxis + f + options + theme_bw() + theme(axis.text.x=element_text(size=6, angle = 90))
}else{
pc.m = pc.m + l + xaxis + yaxis + f + options + theme_bw() + theme(axis.text.x=element_text(size=6, angle = 90))
}
dev.new()
print(pc.m)
if( re.order == TRUE )
{
## of credible intervals, by cluster
datclust.ci = melt(dat.clust, id = c("order","dose","trt","cluster","dose_trtclust"))
datclust.ci$quantile = datclust.ci$variable
datclust.ci$effects = datclust.ci$value
datclust.ci$variable <- datclust.ci$value <- NULL
## although plotting dose_trtclust, set x-axis tick labels to dose (within trt-cluster)
options <- scale_x_discrete(labels=datclust.ci$dose, breaks=datclust.ci$dose_trtclust)
pc.ci = ggplot(data=datclust.ci,aes(x=reorder(dose_trtclust,effects), y = effects, colour = factor(order), shape = factor(order))) + options
}else{
## of credible intervals, by cluster
datclust.ci = melt(dat.clust, id = c("order","dose","trt","cluster"))
datclust.ci$quantile = datclust.ci$variable
datclust.ci$effects = datclust.ci$value
datclust.ci$variable <- datclust.ci$value <- NULL
pc.ci = ggplot(data=datclust.ci,aes(x = dose, y = effects, colour = factor(order), shape = factor(order)))
}
l = geom_line()
l.2 = geom_smooth(aes(group=factor(order)), method = "loess", span = 1.0, alpha = 0.1, se = FALSE)##, colour = "pink")
yaxis = ylab("Effects")
options.2 = labs(colour = "Order", shape="Order")
f = facet_wrap( trt~cluster, scales= "free_x", nrow=nt )
if( smoother == TRUE)
{
pc.ci = pc.ci + l + l.2 + xaxis + yaxis + f + options.2 + theme(axis.text.x=element_text(size=6, angle = 90))
}else{
pc.ci = pc.ci + l + xaxis + yaxis + f + options.2 + theme(axis.text.x=element_text(size=6, angle = 90))
}
dev.new()
print(pc.ci)
if( Nrandom > 1 ) ## more than one effect per dose
{
##
## plot within sample predicted growth curves for each dose, within cluster
##
## set-up structures
C <- length(clusters)
M <- length(trts.plot)
MD <- sum( md )
md.cum <- cumsum( md )
T <- 10 ## length of time intervals
cluster.gc <- trt.gc <- dose.gc <- vector("character", C*MD*T)
y.gc <- time.gc <- dose.order <- vector("numeric", C*MD*T)
X <- object$summary.results$X
measure.times <- unique( X[,(colnames(X) == "time^1")] )
min.T <- min(measure.times); max.T <- max(measure.times)
time <- seq(min.T,max.T,length.out = T)
## compute "response" from effect orders for each dose within treatment within cluster at T timempoints
for( c in 1:C )
{
for( m in 1:M )
{
for( d in 1:md[m] )
{
dat.cmd <- subset( dat.clust, (cluster %in% clusters[c]) & (trt %in% trts.plot[m]) & (dose %in% dose.lab[[m]][d]) )
effect.cmd <- vector("numeric", Nrandom)
for( r in 1:Nrandom )
{
effect.cmd[r] <- dat.cmd$mean.effects[dat.cmd$order == r]
}
for( t in 1:T )
{
## fill return objects
indx.cmdt <- (c-1)*(MD*T) + ((md.cum[m]-md[m])*T) + (d-1)*T + t
cluster.gc[indx.cmdt] <- as.character(clusters[c])
dose.gc[indx.cmdt] <- dose.lab[[m]][d]
dose.order[indx.cmdt] <- dose.ord[[m]][d]
trt.gc[indx.cmdt] <- trts.plot[m]
time.gc[indx.cmdt] <- time[t]
## generate response
y.cmdt <- 0
for( r in 1:Nrandom )
{
y.cmdt <- y.cmdt + effect.cmd[r]*(time[t]^(r-1))
}
y.gc[indx.cmdt] <- y.cmdt
} ## end loop over T time points
} ## end loop over D doses
} ## end loop over M treatments
} ## end loop over C clusters
dat.gc = data.frame(y.gc,time.gc,cluster.gc,trt.gc,dose.gc,dose.order, stringsAsFactors = FALSE)
names(dat.gc) = c("fit","time","cluster","trt","dose","dose.order")
## select doses to plot
dose.plot <- vector("list",nt)
for(m in 1:nt) {dose.plot[[m]] <- sample(dose.lab[[m]],min(length(dose.lab[[m]]),n.dose.plot),replace = FALSE)}
dose.plot <- unlist(dose.plot)
## dose.plot <- c(4,5,6,7,9,10,11,12,16,18,22,23,25,26,29,35,42,45,47,52)
dat.gcplot <- subset(dat.gc, dose %in% dose.plot)
n.clust <- length(unique(dat.gcplot$cluster))
##
## for ALL clients - MEAN growth curve, by TREATMENT category
##
p.gc = ggplot(data=dat.gcplot,aes(x=time, y = fit, colour = factor(reorder(dose,dose.order)))) ## here, looking for individual subject visibility
## l = geom_line(aes(linetype = factor(dose)))
l = geom_line()
l.2 = geom_point()
## l.2 = geom_smooth(aes(group=1),method = "loess", size = 1.1, colour = "black")
axis = labs(x = "Time", y = "Response Units", colour = eval(x.axis.label), linetype = eval(x.axis.label), shape = eval(x.axis.label)) ## x.axis.label captures "dose" label
f = facet_wrap(trt~cluster, ncol = n.clust, scales="fixed")
p.gc = p.gc + l + l.2 + f + axis + scale_colour_grey(start = 0.1, end = 0.5) + theme_bw()
dev.new()
print(p.gc)
}else{ ## Nrandom = 1, so don't plot a growth curve
p.gc <- NULL
dat.gc <- NULL
} ## end conditional statement on Nrandom > 1 to plot by-dose effect growth curves
} ## end conditional statement on whether is.null(map.group)
## return objects
if( is.null(map.group) )
{
if( cred.intervals == FALSE )
{
return(invisible(list(dat.se = dat, p.hm = p.t, p.pp = p.p)))
}else{
return(invisible(list(dat.se = dat, dat.ci = dat.ci, p.hm = p.t, p.pp = p.p)))
}
}else{
if( cred.intervals == FALSE )
{
return(invisible(list(dat.se = dat, dat.clust = dat.clust, dat.clust.ci = datclust.ci, dat.gc = dat.gc, p.hm = p.t, p.pp = p.p, pc.m = pc.m, pc.ci = pc.ci, p.gc = p.gc)))
}else{
return(invisible(list(dat.se = dat, dat.ci = dat.ci, dat.clust = dat.clust, dat.clust.ci = datclust.ci, dat.gc = dat.gc, p.hm = p.t, p.pp = p.p, pc.m = pc.m, pc.ci = pc.ci, p.gc = p.gc)))
}
}
subject <- subjects.num <- label.input <- mean.effects <- dose <- cluster <- fit <- trt <- dose_trtsubject <- dose_trtclust <- effects <- quantile <- NULL
} ## end function ddpEffectsplot
Try the growcurves 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.