\name{cindex.comp.meta} \alias{cindex.comp.meta} %- Also NEED an '\alias' for EACH other topic documented here. \title{Function to compare two concordance indices} \description{ This function compares two lists of concordance indices computed from the same survival data by using the function \code{concordance.index}. The statistical test is a Student t test for dependent samples. } \usage{ cindex.comp.meta(list.cindex1, list.cindex2, hetero = FALSE) } %- maybe also 'usage' for other objects documented here. \arguments{ \item{list.cindex1}{first list of concordance indices as returned by the \code{concordance.index} function.} \item{list.cindex2}{second list of concordance indices as returned by the \code{concordance.index} function.} \item{hetero}{if TRUE, a random effect model is use to compute the meta-estimators. Otherwise a fixed effect model is used.} } \details{ In meta-analysis, we estimate the statistic of interest in several independent datasets. It results a list of estimates such as list of concordance indices. The two lists of concordance indices must be computed from the same samples (and corresponding survival data). The function computes a meta-estimator for the correlations between the two scores and uses a Student t test for dependent samples. } %%\note{ %%Since the comparison between the two concordance indices is done with a paired t-test that relies on the computation of the Pearson correlation coefficient between the original predictors \code{x}, any non-linear transformation of these predictors will affect the computation of the p-value. %%We are now using spearman correlation to circumvent this difficulty. %%} \value{ \item{p.value }{p-value from the Student t test for the comparison cindex1 > cindex2.} \item{cindex1 }{meta-estimator of the first concordance index.} \item{cindex2 }{meta-estimator of the second concordance index.} } \references{Cochrane, W. G. (1954) "The combination of estimates from different experiments", \emph{Biometrics}, \bold{10}, pages 101--129. Haibe-Kains, B. and Desmedt, C. and Sotiriou, C. and Bontempi, G. (2008) "A comparative study of survival models for breast cancer prognostication based on microarray data: does a single gene beat them all?", \emph{Bioinformatics}, \bold{24}, 19, pages 2200--2208.} \author{ Benjamin Haibe-Kains} %\note{ ~~further notes~~ } \seealso{\code{\link{concordance.index}}.} \examples{ #first dataset set.seed(12345) age <- rnorm(100, 50, 10) size <- rexp(100,1) stime <- rexp(100) cens <- runif(100,.5,2) sevent <- as.numeric(stime <= cens) stime <- pmin(stime, cens) c1.1 <- concordance.index(x=age, surv.time=stime, surv.event=sevent, method="noether") c2.1 <- concordance.index(x=size, surv.time=stime, surv.event=sevent, method="noether") #second dataset set.seed(54321) age <- rnorm(110, 53, 10) size <- rexp(110,1.1) stime <- rexp(110) cens <- runif(110,.55,2) sevent <- as.numeric(stime <= cens) stime <- pmin(stime, cens) c1.2 <- concordance.index(x=age, surv.time=stime, surv.event=sevent, method="noether") c2.2 <- concordance.index(x=size, surv.time=stime, surv.event=sevent, method="noether") cindex.comp.meta(list.cindex1=list("cindex.age1"=c1.1, "cindex.age2"=c1.2), list.cindex2=list("cindex.size1"=c2.1, "cindex.size2"=c2.2)) } % Add one or more standard keywords, see file 'KEYWORDS' in the % R documentation directory. \keyword{survival} \keyword{htest}% __ONLY ONE__ keyword per line