Skip to contents

Resampling specifications for evaluate(), tune() and compare()

Source: R/resampling.R
resampling.Rd

These constructors return a lightweight survalis_resampling description that the verb-level functions turn into concrete train/test splits. They do not touch the data themselves, so the same specification can be reused across models to obtain paired comparisons (identical folds via a shared seed).

Usage

cv(v = 5, repeats = 1, strata = "status", seed = NULL)

holdout(prop = 0.75, strata = "status", seed = NULL)

groupcv(group, v = 5, seed = NULL)

bootstrap(times = 25, seed = NULL)

Arguments

v

Integer number of folds (default 5).

repeats

Integer number of times the whole v-fold scheme is repeated with fresh random folds (default 1).

strata

Column name to stratify on, or NA to disable stratification. The default sentinel "status" means "use the event-indicator column of the model formula", whatever its name (for example cens in Surv(time, cens) ~ .). Any other name must be a column of the data; otherwise a warning is given and the resampling is not stratified.

seed

Optional integer seed. Set it (and keep it equal across models) for paired resampling.

prop

Proportion of rows assigned to the training set in holdout() (default 0.75).

group

Column name whose distinct values define the groups held out together in groupcv().

times

Integer number of bootstrap resamples (default 25).

Value

An object of class survalis_resampling.

Details

  • cv() – k-fold cross-validation, optionally repeated, stratified on the event indicator by default.

  • holdout() – a single stratified train/test split.

  • groupcv() – k-fold cross-validation where whole groups (e.g. centres, subjects with repeated rows) are held out together.

  • bootstrap() – resampling with replacement; the out-of-bag rows form the assessment set (Efron-Tibshirani style, suitable for .632 reporting).

Examples

cv(5)
#> <survalis_resampling> 5-fold CV, strata = status
cv(v = 10, repeats = 3, seed = 1)
#> <survalis_resampling> 10-fold CV x 3 repeats, strata = status, seed = 1
holdout(prop = 0.8)
#> <survalis_resampling> holdout (80% train), strata = status
bootstrap(times = 50)
#> <survalis_resampling> bootstrap (50 resamples, OOB assessment)