cv-indices: Create cross-validation folds

cv-indicesR Documentation

Create cross-validation folds


These are helper functions to create cross-validation (CV) folds, i.e., to split up the indices from 1 to n into K subsets ("folds") for K-fold CV. These functions are potentially useful when creating the input for arguments cvfits and cvfun of init_refmodel() (or argument cvfits of cv_varsel.refmodel()). Function cvfolds() is deprecated; please use cv_folds() instead (apart from the name, they are the same). The return value of cv_folds() and cv_ids() is different, see below for details.


cv_folds(n, K, seed = NA)

cvfolds(n, K, seed = NA)

cv_ids(n, K, out = c("foldwise", "indices"), seed = NA)



Number of observations.


Number of folds. Must be at least 2 and not exceed n.


Pseudorandom number generation (PRNG) seed by which the same results can be obtained again if needed. Passed to argument seed of set.seed(), but can also be NA to not call set.seed() at all. If not NA, then the PRNG state is reset (to the state before calling cv_folds() or cv_ids()) upon exiting cv_folds() or cv_ids().


Format of the output, either "foldwise" or "indices". See below for details.


cv_folds() returns a vector of length n such that each element is an integer between 1 and K denoting which fold the corresponding data point belongs to. The return value of cv_ids() depends on the out argument. If out = "foldwise", the return value is a list with K elements, each being a list with elements tr and ts giving the training and test indices, respectively, for the corresponding fold. If out = "indices", the return value is a list with elements tr and ts each being a list with K elements giving the training and test indices, respectively, for each fold.


n <- 100
y <- rnorm(n)
cv <- cv_ids(n, K = 5)
# Mean within the test set of each fold:
cvmeans <- sapply(cv, function(fold) mean(y[fold$ts]))

projpred documentation built on Oct. 1, 2023, 1:07 a.m.