cindex: Computes the concordance index for right censored survival time data

Description

The concordance index (C-index) is the frequency of concordant pairs among all usuable pairs, where a pair consists of the response and a continuous marker of one individual. Without censoring the statistic is related to the Wilcoxon statistic and Somers coefficient. The C-index equals the area under the ROC curve when the response is binary. This function provides the inverse of the probability of censoring weigthing estimate to adjust for right censoring .

Usage

cindex(object,
       data,model
       cens.formula = formula,
       cens.model = c("cox", "marginal", "nonpar", "aalen", "none"),
       conf.int = NULL,
       B = 1000,
       na.action=na.omit,
       tau)

Arguments

object

A named list of prediction models. Each entry is either an R-object for which a predictSurvOrder method exists (see details) or a call that evaluates to such an R-object, see examples.

data

a data frame that contains all the variables in formula and cens.formula

cens.formula

a formula of the form

Surv(time,status) ~
      z1+z2+...

where z1 and z2 are confounders.

cens.model

Method for estimating inverse probability of censoring weigths: cens.model="nonpar" means nonparametric kernel smoothing of the Kaplan-Meier estimator for the censoring variable using symmetric nearest neighborhoods -- this is pos

conf.int

level of confidence: a decimal between 0 and 1, i.e. .95

Number of bootstrap replications

na.action

Passed to model.frame

tau

The upper limit of integration for estimates of restricted C-index. see details section

Value

The C-index

Details

The C-index can be represented by four integrals: two for the event time and two for the continuous marker. The range of the event time variable may be restricted by specifying the upper limit of integration tau.

References

Harrel et al. 1996, Statistics in Medicine, vol 15, pages 361-87