Cross-validated Z-residuals for parametric survival regression models

S3 method for CV.Zresidual() applied to parametric survival regression models fitted with survreg. This method performs K-fold cross-validation to obtain external Z-residuals for model diagnostics.

Usage

# S3 method for class 'survreg'
CV.Zresidual(object, nfolds, foldlist = NULL, data = NULL, nrep = 1, ...)

Arguments

object: A fitted survreg model object.
nfolds: Integer. Number of folds for cross-validation.
foldlist: Optional list specifying custom fold assignments. If NULL, folds are generated internally, typically stratified by the survival response and censoring indicator.
data: Optional data frame used to refit the model during cross-validation. It is highly recommended to supply the original data here to ensure correct model refitting in each fold, especially when the original call was complex.
nrep: Integer. Number of repeated Z-residual samples per observation to generate. Defaults to 1. Each replicate involves re-randomizing the imputed survival probability for censored observations.
...: Further arguments passed to the internal worker function CV_Zresidual_survreg_survival.

Value

An object of class "cvzresid" containing cross-validated Z-residual diagnostics for the parametric survival model. It is a numeric matrix with $N$ rows and nrep columns, accompanied by diagnostic attributes (see CV_Zresidual_survreg_survival for details).

Details

This method delegates the actual cross-validation work to CV_Zresidual_survreg_survival, which handles the iterative refitting of the survreg model on $K-1$ folds and computes the randomized Z-residuals on the held-out fold.

The randomized Z-residual, $Z_{ij}$, for the $j$-th observation in the $i$-th fold is defined as: $$Z_{ij} = ...$$ is computed based on the predicted out-of-sample survival probability $\hat{S}_{\text{train}_i}(t_j)$.

The returned object is tagged with class "cvzresid" in addition to any classes returned by the internal worker.

Examples

if (FALSE) { # \dontrun{
  library(survival)
  # Fit a Weibull model
  fit_weibull <- survreg(Surv(time, status) ~ age + sex,
                         data = lung, dist = "weibull")
  # Compute 5-fold cross-validated Z-residuals
  cv_out <- CV.Zresidual(fit_weibull, nfolds = 5, data = lung, nrep = 10)

  # Check the first few cross-validated residuals
  head(cv_out)
} # }