boot.fun {SemiPar.depCens}

Nonparametric bootstrap approach for the dependent censoring model

Description

This function estimates the bootstrap standard errors for the finite-dimensional model parameters and for the non-parametric cumulative hazard function. Parallel computing using foreach has been used to speed up the estimation of standard errors.

Usage

boot.fun(
  init,
  resData,
  X,
  W,
  lhat,
  cumL,
  dist,
  k,
  lb,
  ub,
  Obs.time,
  cop,
  n.boot,
  n.iter,
  eps
)

Arguments

init

Initial values for the finite dimensional parameters obtained from the fit of fitDepCens

resData

Data matrix with three columns; Z = the observed survival time, d1 = the censoring indicator of T and d2 = the censoring indicator of C.

X

Data matrix with covariates related to T

W

Data matrix with covariates related to C. First column of W should be a vector of ones

lhat

Initial values for the hazard function obtained from the fit of fitDepCens based on the original data.

cumL

Initial values for the cumulative hazard function obtained from the fit of fitDepCens based on the original data.

dist

The distribution to be used for the dependent censoring time C. Only two distributions are allowed, i.e, Weibull and lognormal distributions. With the value "Weibull" as the default.

k

Dimension of X

lb

lower boundary for finite dimensional parameters

ub

Upper boundary for finite dimensional parameters

Obs.time

Observed survival time, which is the minimum of T, C and A, where A is the administrative censoring time.

cop

Which copula should be computed to account for dependency between T and C. This argument can take one of the values from c("Gumbel", "Frank", "Normal").

n.boot

Number of bootstraps to use in the estimation of bootstrap standard errors.

n.iter

Number of iterations; the default is n.iter = 20. The larger the number of iterations, the longer the computational time.

eps

Convergence error. This is set by the user in such away that the desired convergence is met; the default is eps = 1e-3

Value

Bootstrap standard errors for parameter estimates and for estimated cumulative hazard function.