Fold identifiers

Splits observations into balanced and stratified folds.

Usage

.folds(y, family, nfolds)

Arguments

y

response vector of length \(n_0\), containing numerical values (family="gaussian"), integer values (family="poisson"), binary values (family="binomial"), or a survival object created with survival::Surv() (family="cox"), where \(n_0\) is the number of observations used for model training

family

character string "gaussian", "binomial", "poisson", or "cox"

nfolds

positive integer specifying the number of folds (minimum \(3\), maximum \(n\))

NB: If foldid is provided, nfolds is overwritten by max(foldid).

Value

Returns an \(n_0\)-dimensional vector with entries in \(\{1, \ldots, \)nfolds\(\}\).

Details

Randomly splits observations into balanced folds (approximately the same number of observations per fold) and stratified folds (separate splitting for both classes in binomial family or censored/uncensored observations in Cox model).

Examples

# Gaussian and Poisson families
y <- stats::rnorm(n = 100L)
y <- stats::rpois(n = 100L, lambda = 4.0)
foldid <- .folds(y = y, family = "gaussian", nfolds = 10L)
table(foldid)
#> foldid
#>  1  2  3  4  5  6  7  8  9 10 
#> 10 10 10 10 10 10 10 10 10 10 

# binomial family
y <- stats::rbinom(n = 100L, size = 1L, prob = 0.2)
foldid <- .folds(y = y, family = "binomial", nfolds = 10L)
table(y, foldid)
#>    foldid
#> y   1 2 3 4 5 6 7 8 9 10
#>   0 9 9 9 9 9 9 8 8 8  8
#>   1 1 1 1 1 1 1 2 2 2  2

# \donttest{
# Cox model
time <- stats::rexp(n = 100L, rate = 5.0)
status <- stats::rbinom(n = 100L, size = 1L, prob = 0.2)
y <- survival::Surv(time = time, event = status)
foldid <- .folds(y = y, family = "cox", nfolds = 10L)
table(y[, "status"], foldid)
#>    foldid
#>     1 2 3 4 5 6 7 8 9 10
#>   0 8 8 8 8 8 8 8 8 7  7
#>   1 2 2 2 2 2 2 2 2 3  3
# }