fit_surv_option: Options of machine learning methods' wrappers for fitting...

In QIU-Hongxiang-David/SDRsurv: Sequentially Doubly Robust Estimation of Survival Curve with Time-Varying Covariates

Options of machine learning methods' wrappers for fitting conditional survival curves

Description

Options of machine learning methods' wrappers for fitting conditional survival curves

Usage

fit_surv_option(
  nfold = 1,
  option = list(),
  oob = TRUE,
  tune = TRUE,
  tune.option = list(),
  lambda = 0.5
)

Arguments

nfold	number of folds used when fitting survival curves with sample splitting. Default is 1, meaning no sample splitting
option	a list containing optional arguments passed to the wrapped machine learning function. Will be used in a command like do.call(machine.learning, option) where machine.learning is the machine learning function being called. formula and data should not be specified. For randomForestSRC::rfsrc, if tune=TRUE, then mtry and nodesize should not be specified either.
oob	whether to use out-of-bag (OOB) fitted values from random forests (randomForestSRC::rfsrc, party::cforest) and grf::survival_forest) when sample splitting is not used (nfold=1). Ignored otherwise.
tune	whether to tune mtry and nodesize for randomForestSRC::rfsrc. Ignored for other methods.
tune.option	a list containing optional arguments passed to randomForestSRC::tune.rfsrc if randomForestSRC::rfsrc is used and tune=TRUE; ignored otherwise. doBest should not be specified.
lambda	bandwidth parameter for uniform smoothing kernel in nearest neighbours estimation for method "akritas". The default value of 0.5 is arbitrary and should be chosen by the user

QIU-Hongxiang-David/SDRsurv documentation built on March 29, 2024, 8:41 a.m.