R/prcorrlog.R
In Raydonal/prLogistic: Estimation of Prevalence Ratios using Logistic Models
#===================================================================================
# ----------------------------------------------------------------------------------
# ---------- Estimating adjusted prevalence ratio and interval confidence ----------
# ----------------------------------------------------------------------------------
#
# The prcorrlog.R script are a set functions created by
#
# Raydonal Ospina M.
# Department of Statistics
# Federal University of Pernambuco, Recife - Brazil
# contact: rayospina@gmail.com, raydonal@de.ufpe.br
#
# Leila D. Amorim
# Department of Statistics
# Federal University of Bahia, Salvador - Brazil
# contact: leiladen@ufba.br
#
# Originaly created 29 march 2010.
# Last modification: 27 may 2011 at 12:55 Version 1.0
# Last modification: 25 Oct 2011 at 09:10 Version 1.1
# Last modification: 19 Set 2014 at 11:10 Version 1.2
#
# We used some functions to lmer package. In this case, copyright (C)
# Douglas Bates <bates@stat.wisc.edu> and
# Martin Maechler <maechler@R-project.org> All rights reserved.
#
#====================================================================================
#=====================================================================================
# print.cin ( is a method to print IC using classes)
#=====================================================================================
"print.cin" <-
function (x, digits = 5, ...) #max(4, getOption("digits")-3)
{
cat("\n")
cat(x$head, "\n")
rq <- structure(x$ci, names = x$ends)
print(rq, digits = digits)
cat("\n")
invisible(x)
}
#=====================================================================================
#=====================================================================================
# Return the pairs of expressions that separated by vertical bars. ( lmer package )
#=====================================================================================
"findbars" <- function(term)
{
if (is.name(term) || !is.language(term)) return(NULL)
if (term[[1]] == as.name("(")) return(findbars(term[[2]]))
if (!is.call(term)) stop("term must be of class call")
if (term[[1]] == as.name('|')) return(term)
if (length(term) == 2) return(findbars(term[[2]]))
c(findbars(term[[2]]), findbars(term[[3]]))
}
#=====================================================================================
#=====================================================================================
# Return the list of '/'-separated terms in an expression that contains slashes
# ( lmer package )
#=====================================================================================
"slashTerms" <- function(x)
{
if (!("/" %in% all.names(x))) return(x)
if (x[[1]] != as.name("/"))
stop("unparseable formula for grouping factor")
list(slashTerms(x[[2]]), slashTerms(x[[3]]))
}
#=====================================================================================
#=====================================================================================
# from a list of length 2 return recursive interaction terms
# ( lmer package )
#=====================================================================================
"makeInteraction" <- function(x)
{
if (length(x) < 2) return(x)
trm1 <- makeInteraction(x[[1]])
trm11 <- if(is.list(trm1)) trm1[[1]] else trm1
list(substitute(foo:bar, list(foo=x[[2]], bar = trm11)), trm1)
}
#=====================================================================================
#=====================================================================================
# expand any slashes in the grouping factors returned by findbars
# ( lmer package )
#=====================================================================================
"expandSlash" <- function(bb)
{
if (!is.list(bb)) return(expandSlash(list(bb)))
## I really do mean lapply(unlist(... - unlist returns a
## flattened list in this case
unlist(lapply(bb, function(x) {
if (length(x) > 2 && is.list(trms <- slashTerms(x[[3]])))
return(lapply(unlist(makeInteraction(trms)),
function(trm) substitute(foo|bar,
list(foo = x[[2]],
bar = trm))))
x
}))
}
#=====================================================================================
#=====================================================================================
# is.glm (is a glm object?)
#=====================================================================================
"is.glm" <- function (x)
inherits(x, "glm")
#=====================================================================================
#=====================================================================================
# is.glmer ( is a glmer object?)
#=====================================================================================
"is.glmer" <- function (x)
inherits(x, "glmerMod")
#=====================================================================================
#=====================================================================================
# pr.conditional() Estimate prevalence ratios and their confidence intervals using
# logistic models and conditional procedure
#=====================================================================================
"pr.conditional" <- function(object, conf = conf,...)
{
if(is.glm(object))
{
fit.coef = object$coefficients
fit.cov = as.matrix(vcov(object))
m.aux = cbind(1, diag(rep(1,length(fit.coef)-1)))
pr = (1+exp(-fit.coef[1])) / (1+exp(-(m.aux%*%fit.coef)))
p1.hat = exp(fit.coef[1])/(1+exp(fit.coef[1]))
p0.hat = exp((m.aux%*%fit.coef)) / (1+exp((m.aux%*%fit.coef)))
m0.aux= cbind(1, diag(rep(0,length(fit.coef)-1)))
xstart = as.matrix((1-p1.hat)*m0.aux-(diag(as.vector(1-p0.hat)) %*% m.aux))
lnrp = log(pr)
varlogrp = diag((xstart) %*% fit.cov %*% t(xstart))
sdlnrp = sqrt(varlogrp)
quant.cond = abs(qnorm((1-conf)/2))
liminfIC.cond = exp(lnrp - quant.cond*sdlnrp )
limsupIC.cond = exp(lnrp + quant.cond*sdlnrp)
conditional.rp = cbind(pr, liminfIC.cond , limsupIC.cond )
nam = names(fit.coef[-1])
dimnames(conditional.rp) = list(NULL, c("Estimate", paste(as.character(c((1 - conf)/2,
1 - ((1 - conf)/2)) * 100), "%", sep = "")))
rownames(conditional.rp) = nam
conditional.rp
}
else
{
if(is.glmer(object))
{
fit.coef= object@beta
fit.cov = as.matrix(vcov(object))
m.aux = cbind(1, diag(rep(1,length(fit.coef)-1)))
pr = (1+exp(-fit.coef[1])) / (1+exp(-(m.aux%*%fit.coef)))
p1.hat = exp(fit.coef[1])/(1+exp(fit.coef[1]))
p0.hat = exp((m.aux%*%fit.coef)) / (1+exp((m.aux%*%fit.coef)))
m0.aux= cbind(1, diag(rep(0,length(fit.coef)-1)))
xstart = as.matrix((1-p1.hat)*m0.aux-(diag(as.vector(1-p0.hat)) %*% m.aux))
lnrp = log(pr)
varlogrp = diag((xstart) %*% fit.cov %*% t(xstart))
sdlnrp = sqrt(varlogrp)
quant.cond = abs(qnorm((1-conf)/2))
liminfIC.cond = exp(lnrp - quant.cond*sdlnrp )
liminfIC.cond = ifelse(liminfIC.cond< 0, 0, liminfIC.cond) ## Truncate the IC for negative values
limsupIC.cond = exp(lnrp + quant.cond*sdlnrp)
limsupIC.cond = ifelse(limsupIC.cond< 0, 0, limsupIC.cond) ## Truncate the IC for negative values
conditional.rp = cbind(pr, liminfIC.cond , limsupIC.cond )
nam = names(fit.coef[-1])
dimnames(conditional.rp) = list(NULL, c("Estimate", paste(as.character(c((1 - conf)/2,
1 - ((1 - conf)/2)) * 100), "%", sep = "")))
rownames(conditional.rp) = nam
conditional.rp
}
else
{
stop("There are not glm and glmer models. Revise arguments and try again")
}
}
conditional.rp
}
#=====================================================================================
#=====================================================================================
# pr.marginal() Estimate prevalence ratios and their confidence intervals using
# logistic models and marginal procedure
#=====================================================================================
"pr.marginal" <- function(object, conf = conf,...)
{
if(is.glm(object))
{
#glm class model
coef.glm = object$coefficients
fit.coef = as.vector(t(coef.glm))
fit.cov = as.matrix(vcov(object))
m.aux = as.matrix(object$model[-1])
m.covariate = as.matrix(cbind(1,m.aux))
n.coef = dim(m.covariate)[2]
p1.marg.aux = colMeans(m.aux)
p0.marg.aux = 1- p1.marg.aux #mean(m.aux)
marginal.rp = NULL
m.ind0 = cbind(1, diag(1,(n.coef-1)))
m.ind1 = cbind(1, diag(0,(n.coef-1)))
for(i in 1:(n.coef-1))
{
m.aux.p0 = m.covariate
m.aux.p0[,i+1] = 0
p0.marg = c(1,p0.marg.aux)
p0.marg[i+1]=0
p0.aux = (exp(m.aux.p0%*%fit.coef))/(1+exp(m.aux.p0%*%fit.coef))
p0.med = mean(p0.aux)
m.aux.p1 = m.covariate
m.aux.p1[,i+1] = 1
p1.marg = c(1,p1.marg.aux)
p1.marg[i+1]=0
p1.aux = (exp(m.aux.p1%*%fit.coef))/(1+exp(m.aux.p1%*%fit.coef))
p1.med = mean(p1.aux)
rp.marg = p1.med / p0.med #prevalence ratio
# IC by delta method
lnrp.marg = log(rp.marg)
# xstar.marg = (1-p1.med)*p1.marg - (1-p0.med)*p0.marg
xstar.marg = ((1-p1.med) * m.ind0[i,]) - ((1-p0.med)*m.ind1[i,])
varlogrp.marg = as.vector(xstar.marg) %*% as.matrix(fit.cov) %*% as.vector(t(xstar.marg))
stdlogrp.marg = sqrt(varlogrp.marg)
quant.marg = abs(qnorm((1-conf)/2))
liminfIC.marg = exp(lnrp.marg - quant.marg*stdlogrp.marg )
liminfIC.marg = ifelse(liminfIC.marg< 0, 0, liminfIC.marg) ## Truncate the IC for negative values
limsupIC.marg = exp(lnrp.marg + quant.marg*stdlogrp.marg )
limsupIC.marg = ifelse(limsupIC.marg< 0, 0, limsupIC.marg) ## Truncate the IC for negative values
marginal.rp =(cbind(marginal.rp, c(rp.marg, liminfIC.marg, limsupIC.marg) ) ) # pr whit IC
}
marginal.rp = t(marginal.rp)
nam = names(coef.glm[-1])
dimnames(marginal.rp) = list(NULL, c("Estimate", paste(as.character(c((1 - conf)/2,
1 - ((1 - conf)/2)) * 100), "%", sep = "")))
rownames(marginal.rp) = nam
marginal.rp # output relative prevalence whit IC by delta method
}
else
{
# glmer class model
coef.lmer = object@beta
fit.coef = as.vector(t(coef.lmer))
fit.cov = as.matrix(vcov(object))
m.model = object@frame
m.aux = as.matrix(m.model[c(-1, -dim(m.model)[2])])
m.covariate = as.matrix(cbind(1,m.aux))
n.coef = dim(m.covariate)[2]
p1.marg.aux = colMeans(m.aux)
p0.marg.aux = 1- p1.marg.aux #mean(m.aux)
marginal.rp = NULL
m.ind0 = cbind(1, diag(1,(n.coef-1)))
m.ind1 = cbind(1, diag(0,(n.coef-1)))
for(i in 1:(n.coef-1))
{
m.aux.p0 = m.covariate
m.aux.p0[,i+1] = 0
p0.marg = c(1,p0.marg.aux)
p0.marg[i+1]=0
p0.aux = (exp(m.aux.p0%*%fit.coef))/(1+exp(m.aux.p0%*%fit.coef))
p0.med = mean(p0.aux)
m.aux.p1 = m.covariate
m.aux.p1[,i+1] = 1
p1.marg = c(1,p1.marg.aux)
p1.marg[i+1]=0
p1.aux = (exp(m.aux.p1%*%fit.coef))/(1+exp(m.aux.p1%*%fit.coef))
p1.med = mean(p1.aux)
rp.marg = p1.med / p0.med #prevalence ratio
# IC by delta method
lnrp.marg = log(rp.marg)
xstar.marg = ((1-p1.med) * m.ind0[i,]) - ((1-p0.med)*m.ind1[i,])
#xstar.marg = (1-p1.med)*p1.marg - (1-p0.med)*p0.marg
varlogrp.marg = as.vector(xstar.marg) %*% as.matrix(fit.cov) %*% as.vector(t(xstar.marg))
stdlogrp.marg = sqrt(varlogrp.marg)
quant.marg = abs(qnorm((1-conf)/2))
liminfIC.marg = exp(lnrp.marg - quant.marg*stdlogrp.marg )
limsupIC.marg = exp(lnrp.marg + quant.marg*stdlogrp.marg )
marginal.rp =(cbind(marginal.rp, c(rp.marg,liminfIC.marg, limsupIC.marg) ) ) # pr whit IC
}
marginal.rp = t(marginal.rp)
nam = names(coef.lmer[-1])
dimnames(marginal.rp) = list(NULL, c("Estimate", paste(as.character(c((1 - conf)/2,
1 - ((1 - conf)/2)) * 100), "%", sep = "")))
rownames(marginal.rp) = nam
marginal.rp # output relative prevalence whit IC by delta method
}
}
#=====================================================================================
#=====================================================================================
# boot.conditional() Auxiliar bootstrap function to calculate confidence intervals to
# prevalence ratio using the conditional model
#=====================================================================================
"boot.conditional" <- function(data, indices, object, conf,...){
data<-data[indices,]
fit = update(object, data=data)
obj.aux = pr.conditional(fit,conf)
as.vector(t(obj.aux[,1]))
}
#=====================================================================================
#=====================================================================================
# boot.marg() Auxiliar bootstrap function to calculate confidence intervals to
# prevalence ratio using the marginal model
#=====================================================================================
"boot.marg" <- function(data, indices, object, conf,...){
data<-data[indices,]
fit = update(object, data=data)
obj.aux = pr.marginal(fit,conf)
as.vector(t(obj.aux[,1]))
}
#=====================================================================================
#=====================================================================================
# prLogisticBootMarg() Confidence intervals to prevalence ratio using the marginal model
# by default the Bootstrap IC are obtained by using the normal and percentile method
# whit confidence nivel at 95 %
#=====================================================================================
prLogisticBootMarg <- function(
object, # A fitted model
data, # data set used by estimate the model
conf = 0.95, # A scalar or vector containing the confidence level(s) of the required interval(s).
R = 99, # The number of bootstrap replicates
... # ... Any extra arguments
)
{
#require(boot)
if(is.glm(object))
{
nam = names(object$coefficients[-1])
}
else
{
if(is.glmer(object))
{
nam = names(object@beta[-1])
}
else{
stop("There are not glm and glmer models. Review the arguments function and try again")
}
}
output= boot(data = data, statistic = boot.marg, R = R, object=object, conf=conf,...)
n = length(output$t0)
ic.boot = NULL
for(i in 1:n)
{
ic.boot.aux = boot.ci(output, type=c("norm", "perc"),index=i)
ic.boot.normal = ic.boot.aux$normal[c(2,3)]
ic.boot.percent = ic.boot.aux$percent[c(4,5)]
ic.boot = cbind(ic.boot, c(ic.boot.normal, ic.boot.percent))
}
ic.boot = ifelse(ic.boot< 0, 0, ic.boot) ## Truncate the IC for negatve values
ic.boot=cbind(output$t0,t(ic.boot))
dimnames(ic.boot) = list(NULL, c("Estimate", paste(as.character(c(
(1 - conf)/2, 1 - ((1 - conf)/2), (1 - conf)/2, 1 - ((1 - conf)/2)) * 100),
"%", sep = "")))
rownames(ic.boot) = nam
ic.boot # We can modified this function to obtain other statistics
res <- list() # Output
head1 <- paste(paste(as.character(conf * 100), "%", sep = ""),
c("Confidence Interval using Bootstrap Method\n"))
head2 <- c("\t\t Normal\t Percentile")
res$head <- rbind(head1, head2)
res$ci = ic.boot
class(res) <- "cin"
res
}
#=====================================================================================
#=====================================================================================
# prLogisticBootCond() Confidence intervals to prevalence ratio using the conditional model
# by default the Bootstrap IC are obtained by using the normal and percentile method
# whit confidence nivel at 95 %
#=====================================================================================
prLogisticBootCond =function(
object, # A fitted model
data, # data set used by estimate the model
conf = 0.95, # A scalar or vector containing the confidence level(s) of the required interval(s).
R = 99, # The number of bootstrap replicates
... # ... Any extra arguments
)
{
#require(boot)
if(is.glm(object))
{
nam = names(object$coefficients[-1])
}
else
{
if(is.glmer(object))
{
nam = names(object@beta[-1])
}
else{
stop("There are not glm and glmer models. Review the arguments functions and try again")
}
}
output= boot(data = data, statistic = boot.conditional, R = R, object=object, conf=conf,...)
n = length(output$t0)
ic.boot = NULL
for(i in 1:n)
{
ic.boot.aux = boot.ci(output, type=c("norm", "perc"),index=i)
ic.boot.normal = ic.boot.aux$normal[c(2,3)]
ic.boot.percent = ic.boot.aux$percent[c(4,5)]
ic.boot = cbind(ic.boot, c(ic.boot.normal, ic.boot.percent))
}
ic.boot=cbind(output$t0,t(ic.boot))
ic.boot = ifelse(ic.boot< 0, 0, ic.boot) ## Truncate the IC for negatve values
dimnames(ic.boot) = list(NULL, c("Estimate", paste(as.character(c(
(1 - conf)/2, 1 - ((1 - conf)/2), (1 - conf)/2, 1 - ((1 - conf)/2)) * 100),
"%", sep = "")))
rownames(ic.boot) = nam
ic.boot # We can modified this function to obtain other statistics
res <- list() # Output
head1 <- paste(paste(as.character(conf * 100), "%", sep = ""),
c("Confidence Interval using Bootstrap Method\n"))
head2 <- c("\t\t Normal\t Percentile")
res$head <- rbind(head1, head2)
res$ci = ic.boot
class(res) <- "cin"
res
}
#=====================================================================================
#=====================================================================================
# prLogisticDelta() Estimate prevalence ratios and their confidence intervals using
# logistic models
#=====================================================================================
prLogisticDelta <- function (
formula, # A symbolic description of the model to be fitted.
cluster = FALSE, # A logical value indicating if the model have a random effects
pattern = c("conditional", "marginal"), # Method for estimate prevalence ratio
conf = 0.95, # A scalar or vector containing the confidence level(s) of the required interval(s).
dataset, # data set used to estimate the model
... # ... Any extra arguments
)
{
#require(lme4)
if (!cluster)
{
bars <- expandSlash(findbars(formula[[3]]))
if ((!length(bars)==FALSE))
{
stop("Random effects terms specified in formula. Use cluster = TRUE ")
}
else
{
fit = glm(formula, family=binomial, data=dataset)
}
}
else
{
if (cluster)
bars <- expandSlash(findbars(formula[[3]]))
if (!length(bars))
{
stop("No random effects terms specified in formula. Use cluster = FALSE")
}
else
{
fit = glmer(formula, family = binomial, data = dataset)
}
}
fit
pattern <- match.arg(pattern)
if((pattern=="conditional") )
{
out = pr.conditional(fit, conf=conf,...)
}
else
{
if((pattern=="marginal") )
{
out = pr.marginal(fit, conf=conf,...)
}
else
{
stop("There are some problem whit the arguments. Review and try again")
}
}
res <- list()
res$head <- paste(paste(as.character(conf * 100), "%", sep = ""),
c("Confidence Interval using Delta method"))
res$ci <- out
class(res) <- "cin"
res
}
#=====================================================================================
Raydonal/prLogistic documentation built on May 28, 2019, 2:27 p.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.
#===================================================================================
# ----------------------------------------------------------------------------------
# ---------- Estimating adjusted prevalence ratio and interval confidence ----------
# ----------------------------------------------------------------------------------
#
# The prcorrlog.R script are a set functions created by
#
# Raydonal Ospina M.
# Department of Statistics
# Federal University of Pernambuco, Recife - Brazil
# contact: rayospina@gmail.com, raydonal@de.ufpe.br
#
# Leila D. Amorim
# Department of Statistics
# Federal University of Bahia, Salvador - Brazil
# contact: leiladen@ufba.br
#
# Originaly created 29 march 2010.
# Last modification: 27 may 2011 at 12:55 Version 1.0
# Last modification: 25 Oct 2011 at 09:10 Version 1.1
# Last modification: 19 Set 2014 at 11:10 Version 1.2
#
# We used some functions to lmer package. In this case, copyright (C)
# Douglas Bates <bates@stat.wisc.edu> and
# Martin Maechler <maechler@R-project.org> All rights reserved.
#
#====================================================================================
#=====================================================================================
# print.cin ( is a method to print IC using classes)
#=====================================================================================
"print.cin" <-
function (x, digits = 5, ...) #max(4, getOption("digits")-3)
{
cat("\n")
cat(x$head, "\n")
rq <- structure(x$ci, names = x$ends)
print(rq, digits = digits)
cat("\n")
invisible(x)
}
#=====================================================================================
#=====================================================================================
# Return the pairs of expressions that separated by vertical bars. ( lmer package )
#=====================================================================================
"findbars" <- function(term)
{
if (is.name(term) || !is.language(term)) return(NULL)
if (term[[1]] == as.name("(")) return(findbars(term[[2]]))
if (!is.call(term)) stop("term must be of class call")
if (term[[1]] == as.name('|')) return(term)
if (length(term) == 2) return(findbars(term[[2]]))
c(findbars(term[[2]]), findbars(term[[3]]))
}
#=====================================================================================
#=====================================================================================
# Return the list of '/'-separated terms in an expression that contains slashes
# ( lmer package )
#=====================================================================================
"slashTerms" <- function(x)
{
if (!("/" %in% all.names(x))) return(x)
if (x[[1]] != as.name("/"))
stop("unparseable formula for grouping factor")
list(slashTerms(x[[2]]), slashTerms(x[[3]]))
}
#=====================================================================================
#=====================================================================================
# from a list of length 2 return recursive interaction terms
# ( lmer package )
#=====================================================================================
"makeInteraction" <- function(x)
{
if (length(x) < 2) return(x)
trm1 <- makeInteraction(x[[1]])
trm11 <- if(is.list(trm1)) trm1[[1]] else trm1
list(substitute(foo:bar, list(foo=x[[2]], bar = trm11)), trm1)
}
#=====================================================================================
#=====================================================================================
# expand any slashes in the grouping factors returned by findbars
# ( lmer package )
#=====================================================================================
"expandSlash" <- function(bb)
{
if (!is.list(bb)) return(expandSlash(list(bb)))
## I really do mean lapply(unlist(... - unlist returns a
## flattened list in this case
unlist(lapply(bb, function(x) {
if (length(x) > 2 && is.list(trms <- slashTerms(x[[3]])))
return(lapply(unlist(makeInteraction(trms)),
function(trm) substitute(foo|bar,
list(foo = x[[2]],
bar = trm))))
x
}))
}
#=====================================================================================
#=====================================================================================
# is.glm (is a glm object?)
#=====================================================================================
"is.glm" <- function (x)
inherits(x, "glm")
#=====================================================================================
#=====================================================================================
# is.glmer ( is a glmer object?)
#=====================================================================================
"is.glmer" <- function (x)
inherits(x, "glmerMod")
#=====================================================================================
#=====================================================================================
# pr.conditional() Estimate prevalence ratios and their confidence intervals using
# logistic models and conditional procedure
#=====================================================================================
"pr.conditional" <- function(object, conf = conf,...)
{
if(is.glm(object))
{
fit.coef = object$coefficients
fit.cov = as.matrix(vcov(object))
m.aux = cbind(1, diag(rep(1,length(fit.coef)-1)))
pr = (1+exp(-fit.coef[1])) / (1+exp(-(m.aux%*%fit.coef)))
p1.hat = exp(fit.coef[1])/(1+exp(fit.coef[1]))
p0.hat = exp((m.aux%*%fit.coef)) / (1+exp((m.aux%*%fit.coef)))
m0.aux= cbind(1, diag(rep(0,length(fit.coef)-1)))
xstart = as.matrix((1-p1.hat)*m0.aux-(diag(as.vector(1-p0.hat)) %*% m.aux))
lnrp = log(pr)
varlogrp = diag((xstart) %*% fit.cov %*% t(xstart))
sdlnrp = sqrt(varlogrp)
quant.cond = abs(qnorm((1-conf)/2))
liminfIC.cond = exp(lnrp - quant.cond*sdlnrp )
limsupIC.cond = exp(lnrp + quant.cond*sdlnrp)
conditional.rp = cbind(pr, liminfIC.cond , limsupIC.cond )
nam = names(fit.coef[-1])
dimnames(conditional.rp) = list(NULL, c("Estimate", paste(as.character(c((1 - conf)/2,
1 - ((1 - conf)/2)) * 100), "%", sep = "")))
rownames(conditional.rp) = nam
conditional.rp
}
else
{
if(is.glmer(object))
{
fit.coef= object@beta
fit.cov = as.matrix(vcov(object))
m.aux = cbind(1, diag(rep(1,length(fit.coef)-1)))
pr = (1+exp(-fit.coef[1])) / (1+exp(-(m.aux%*%fit.coef)))
p1.hat = exp(fit.coef[1])/(1+exp(fit.coef[1]))
p0.hat = exp((m.aux%*%fit.coef)) / (1+exp((m.aux%*%fit.coef)))
m0.aux= cbind(1, diag(rep(0,length(fit.coef)-1)))
xstart = as.matrix((1-p1.hat)*m0.aux-(diag(as.vector(1-p0.hat)) %*% m.aux))
lnrp = log(pr)
varlogrp = diag((xstart) %*% fit.cov %*% t(xstart))
sdlnrp = sqrt(varlogrp)
quant.cond = abs(qnorm((1-conf)/2))
liminfIC.cond = exp(lnrp - quant.cond*sdlnrp )
liminfIC.cond = ifelse(liminfIC.cond< 0, 0, liminfIC.cond) ## Truncate the IC for negative values
limsupIC.cond = exp(lnrp + quant.cond*sdlnrp)
limsupIC.cond = ifelse(limsupIC.cond< 0, 0, limsupIC.cond) ## Truncate the IC for negative values
conditional.rp = cbind(pr, liminfIC.cond , limsupIC.cond )
nam = names(fit.coef[-1])
dimnames(conditional.rp) = list(NULL, c("Estimate", paste(as.character(c((1 - conf)/2,
1 - ((1 - conf)/2)) * 100), "%", sep = "")))
rownames(conditional.rp) = nam
conditional.rp
}
else
{
stop("There are not glm and glmer models. Revise arguments and try again")
}
}
conditional.rp
}
#=====================================================================================
#=====================================================================================
# pr.marginal() Estimate prevalence ratios and their confidence intervals using
# logistic models and marginal procedure
#=====================================================================================
"pr.marginal" <- function(object, conf = conf,...)
{
if(is.glm(object))
{
#glm class model
coef.glm = object$coefficients
fit.coef = as.vector(t(coef.glm))
fit.cov = as.matrix(vcov(object))
m.aux = as.matrix(object$model[-1])
m.covariate = as.matrix(cbind(1,m.aux))
n.coef = dim(m.covariate)[2]
p1.marg.aux = colMeans(m.aux)
p0.marg.aux = 1- p1.marg.aux #mean(m.aux)
marginal.rp = NULL
m.ind0 = cbind(1, diag(1,(n.coef-1)))
m.ind1 = cbind(1, diag(0,(n.coef-1)))
for(i in 1:(n.coef-1))
{
m.aux.p0 = m.covariate
m.aux.p0[,i+1] = 0
p0.marg = c(1,p0.marg.aux)
p0.marg[i+1]=0
p0.aux = (exp(m.aux.p0%*%fit.coef))/(1+exp(m.aux.p0%*%fit.coef))
p0.med = mean(p0.aux)
m.aux.p1 = m.covariate
m.aux.p1[,i+1] = 1
p1.marg = c(1,p1.marg.aux)
p1.marg[i+1]=0
p1.aux = (exp(m.aux.p1%*%fit.coef))/(1+exp(m.aux.p1%*%fit.coef))
p1.med = mean(p1.aux)
rp.marg = p1.med / p0.med #prevalence ratio
# IC by delta method
lnrp.marg = log(rp.marg)
# xstar.marg = (1-p1.med)*p1.marg - (1-p0.med)*p0.marg
xstar.marg = ((1-p1.med) * m.ind0[i,]) - ((1-p0.med)*m.ind1[i,])
varlogrp.marg = as.vector(xstar.marg) %*% as.matrix(fit.cov) %*% as.vector(t(xstar.marg))
stdlogrp.marg = sqrt(varlogrp.marg)
quant.marg = abs(qnorm((1-conf)/2))
liminfIC.marg = exp(lnrp.marg - quant.marg*stdlogrp.marg )
liminfIC.marg = ifelse(liminfIC.marg< 0, 0, liminfIC.marg) ## Truncate the IC for negative values
limsupIC.marg = exp(lnrp.marg + quant.marg*stdlogrp.marg )
limsupIC.marg = ifelse(limsupIC.marg< 0, 0, limsupIC.marg) ## Truncate the IC for negative values
marginal.rp =(cbind(marginal.rp, c(rp.marg, liminfIC.marg, limsupIC.marg) ) ) # pr whit IC
}
marginal.rp = t(marginal.rp)
nam = names(coef.glm[-1])
dimnames(marginal.rp) = list(NULL, c("Estimate", paste(as.character(c((1 - conf)/2,
1 - ((1 - conf)/2)) * 100), "%", sep = "")))
rownames(marginal.rp) = nam
marginal.rp # output relative prevalence whit IC by delta method
}
else
{
# glmer class model
coef.lmer = object@beta
fit.coef = as.vector(t(coef.lmer))
fit.cov = as.matrix(vcov(object))
m.model = object@frame
m.aux = as.matrix(m.model[c(-1, -dim(m.model)[2])])
m.covariate = as.matrix(cbind(1,m.aux))
n.coef = dim(m.covariate)[2]
p1.marg.aux = colMeans(m.aux)
p0.marg.aux = 1- p1.marg.aux #mean(m.aux)
marginal.rp = NULL
m.ind0 = cbind(1, diag(1,(n.coef-1)))
m.ind1 = cbind(1, diag(0,(n.coef-1)))
for(i in 1:(n.coef-1))
{
m.aux.p0 = m.covariate
m.aux.p0[,i+1] = 0
p0.marg = c(1,p0.marg.aux)
p0.marg[i+1]=0
p0.aux = (exp(m.aux.p0%*%fit.coef))/(1+exp(m.aux.p0%*%fit.coef))
p0.med = mean(p0.aux)
m.aux.p1 = m.covariate
m.aux.p1[,i+1] = 1
p1.marg = c(1,p1.marg.aux)
p1.marg[i+1]=0
p1.aux = (exp(m.aux.p1%*%fit.coef))/(1+exp(m.aux.p1%*%fit.coef))
p1.med = mean(p1.aux)
rp.marg = p1.med / p0.med #prevalence ratio
# IC by delta method
lnrp.marg = log(rp.marg)
xstar.marg = ((1-p1.med) * m.ind0[i,]) - ((1-p0.med)*m.ind1[i,])
#xstar.marg = (1-p1.med)*p1.marg - (1-p0.med)*p0.marg
varlogrp.marg = as.vector(xstar.marg) %*% as.matrix(fit.cov) %*% as.vector(t(xstar.marg))
stdlogrp.marg = sqrt(varlogrp.marg)
quant.marg = abs(qnorm((1-conf)/2))
liminfIC.marg = exp(lnrp.marg - quant.marg*stdlogrp.marg )
limsupIC.marg = exp(lnrp.marg + quant.marg*stdlogrp.marg )
marginal.rp =(cbind(marginal.rp, c(rp.marg,liminfIC.marg, limsupIC.marg) ) ) # pr whit IC
}
marginal.rp = t(marginal.rp)
nam = names(coef.lmer[-1])
dimnames(marginal.rp) = list(NULL, c("Estimate", paste(as.character(c((1 - conf)/2,
1 - ((1 - conf)/2)) * 100), "%", sep = "")))
rownames(marginal.rp) = nam
marginal.rp # output relative prevalence whit IC by delta method
}
}
#=====================================================================================
#=====================================================================================
# boot.conditional() Auxiliar bootstrap function to calculate confidence intervals to
# prevalence ratio using the conditional model
#=====================================================================================
"boot.conditional" <- function(data, indices, object, conf,...){
data<-data[indices,]
fit = update(object, data=data)
obj.aux = pr.conditional(fit,conf)
as.vector(t(obj.aux[,1]))
}
#=====================================================================================
#=====================================================================================
# boot.marg() Auxiliar bootstrap function to calculate confidence intervals to
# prevalence ratio using the marginal model
#=====================================================================================
"boot.marg" <- function(data, indices, object, conf,...){
data<-data[indices,]
fit = update(object, data=data)
obj.aux = pr.marginal(fit,conf)
as.vector(t(obj.aux[,1]))
}
#=====================================================================================
#=====================================================================================
# prLogisticBootMarg() Confidence intervals to prevalence ratio using the marginal model
# by default the Bootstrap IC are obtained by using the normal and percentile method
# whit confidence nivel at 95 %
#=====================================================================================
prLogisticBootMarg <- function(
object, # A fitted model
data, # data set used by estimate the model
conf = 0.95, # A scalar or vector containing the confidence level(s) of the required interval(s).
R = 99, # The number of bootstrap replicates
... # ... Any extra arguments
)
{
#require(boot)
if(is.glm(object))
{
nam = names(object$coefficients[-1])
}
else
{
if(is.glmer(object))
{
nam = names(object@beta[-1])
}
else{
stop("There are not glm and glmer models. Review the arguments function and try again")
}
}
output= boot(data = data, statistic = boot.marg, R = R, object=object, conf=conf,...)
n = length(output$t0)
ic.boot = NULL
for(i in 1:n)
{
ic.boot.aux = boot.ci(output, type=c("norm", "perc"),index=i)
ic.boot.normal = ic.boot.aux$normal[c(2,3)]
ic.boot.percent = ic.boot.aux$percent[c(4,5)]
ic.boot = cbind(ic.boot, c(ic.boot.normal, ic.boot.percent))
}
ic.boot = ifelse(ic.boot< 0, 0, ic.boot) ## Truncate the IC for negatve values
ic.boot=cbind(output$t0,t(ic.boot))
dimnames(ic.boot) = list(NULL, c("Estimate", paste(as.character(c(
(1 - conf)/2, 1 - ((1 - conf)/2), (1 - conf)/2, 1 - ((1 - conf)/2)) * 100),
"%", sep = "")))
rownames(ic.boot) = nam
ic.boot # We can modified this function to obtain other statistics
res <- list() # Output
head1 <- paste(paste(as.character(conf * 100), "%", sep = ""),
c("Confidence Interval using Bootstrap Method\n"))
head2 <- c("\t\t Normal\t Percentile")
res$head <- rbind(head1, head2)
res$ci = ic.boot
class(res) <- "cin"
res
}
#=====================================================================================
#=====================================================================================
# prLogisticBootCond() Confidence intervals to prevalence ratio using the conditional model
# by default the Bootstrap IC are obtained by using the normal and percentile method
# whit confidence nivel at 95 %
#=====================================================================================
prLogisticBootCond =function(
object, # A fitted model
data, # data set used by estimate the model
conf = 0.95, # A scalar or vector containing the confidence level(s) of the required interval(s).
R = 99, # The number of bootstrap replicates
... # ... Any extra arguments
)
{
#require(boot)
if(is.glm(object))
{
nam = names(object$coefficients[-1])
}
else
{
if(is.glmer(object))
{
nam = names(object@beta[-1])
}
else{
stop("There are not glm and glmer models. Review the arguments functions and try again")
}
}
output= boot(data = data, statistic = boot.conditional, R = R, object=object, conf=conf,...)
n = length(output$t0)
ic.boot = NULL
for(i in 1:n)
{
ic.boot.aux = boot.ci(output, type=c("norm", "perc"),index=i)
ic.boot.normal = ic.boot.aux$normal[c(2,3)]
ic.boot.percent = ic.boot.aux$percent[c(4,5)]
ic.boot = cbind(ic.boot, c(ic.boot.normal, ic.boot.percent))
}
ic.boot=cbind(output$t0,t(ic.boot))
ic.boot = ifelse(ic.boot< 0, 0, ic.boot) ## Truncate the IC for negatve values
dimnames(ic.boot) = list(NULL, c("Estimate", paste(as.character(c(
(1 - conf)/2, 1 - ((1 - conf)/2), (1 - conf)/2, 1 - ((1 - conf)/2)) * 100),
"%", sep = "")))
rownames(ic.boot) = nam
ic.boot # We can modified this function to obtain other statistics
res <- list() # Output
head1 <- paste(paste(as.character(conf * 100), "%", sep = ""),
c("Confidence Interval using Bootstrap Method\n"))
head2 <- c("\t\t Normal\t Percentile")
res$head <- rbind(head1, head2)
res$ci = ic.boot
class(res) <- "cin"
res
}
#=====================================================================================
#=====================================================================================
# prLogisticDelta() Estimate prevalence ratios and their confidence intervals using
# logistic models
#=====================================================================================
prLogisticDelta <- function (
formula, # A symbolic description of the model to be fitted.
cluster = FALSE, # A logical value indicating if the model have a random effects
pattern = c("conditional", "marginal"), # Method for estimate prevalence ratio
conf = 0.95, # A scalar or vector containing the confidence level(s) of the required interval(s).
dataset, # data set used to estimate the model
... # ... Any extra arguments
)
{
#require(lme4)
if (!cluster)
{
bars <- expandSlash(findbars(formula[[3]]))
if ((!length(bars)==FALSE))
{
stop("Random effects terms specified in formula. Use cluster = TRUE ")
}
else
{
fit = glm(formula, family=binomial, data=dataset)
}
}
else
{
if (cluster)
bars <- expandSlash(findbars(formula[[3]]))
if (!length(bars))
{
stop("No random effects terms specified in formula. Use cluster = FALSE")
}
else
{
fit = glmer(formula, family = binomial, data = dataset)
}
}
fit
pattern <- match.arg(pattern)
if((pattern=="conditional") )
{
out = pr.conditional(fit, conf=conf,...)
}
else
{
if((pattern=="marginal") )
{
out = pr.marginal(fit, conf=conf,...)
}
else
{
stop("There are some problem whit the arguments. Review and try again")
}
}
res <- list()
res$head <- paste(paste(as.character(conf * 100), "%", sep = ""),
c("Confidence Interval using Delta method"))
res$ci <- out
class(res) <- "cin"
res
}
#=====================================================================================
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.