complexity.ipec.rsf_mtry {peperr}R Documentation

Interface function for complexity selection for random survival forest via integrated prediction error curve and the bootstrap

Description

Determines the optimal value for the number of candidate values at each split for a survival model fitted by random survival forest via integrated prediction error curve (IPEC) estimates, conforming to the calling convention required by argument complexity in peperr call.

Usage

complexity.ipec.rsf_mtry(response, x, boot.n.c = 10, mtry, 
   eval.times = NULL, full.data, relative=TRUE, ...)

Arguments

response a survival object (with Surv(time, status)).
x n*p matrix of covariates.
boot.n.c number of bootstrap samples.
mtry vector of potential values for parameter mtry, i.e. number of randomly selected variables at each split.
eval.times vector of evaluation time points.
full.data Data frame containing response and covariates of the full data set.
relative Should relative IPEC be used for evaluation?
... additional arguments passed to rsf call.

Details

Plotting the .632+ estimator for each time point given in eval.times results in a prediction error curve. A summary measure can be obtained by integrating over time. complexity.ipec.rsf_mtry uses a Lebesgue-like integral taking Kaplan-Meier estimates as weights. The optimal value for parameter mtry out of the set of passed values, for which the minimal IPEC is obtained, is returned.

Value

Scalar value giving the optimal value for parameter mtry.

See Also

peperr, rsf


[Package peperr version 1.1-5 Index]