Skip to contents

Quantile Bins with an Explicit Missing Level

Source: R/binned_missing_cox.R
cut_with_missing_category.Rd

Cuts a numeric variable into equal-count bins and gives missing values their own level, so a survival model can ask whether being unmeasured carries risk of its own. The reference level comes first (what `coxph()` and `lm()` treat as the baseline) and the missing level last.

Usage

cut_with_missing_category(
  x,
  n_bins = 3,
  labels = NULL,
  missing_label = "Missing",
  reference = NULL,
  subset_mask = NULL
)

Arguments

x

Numeric vector.

n_bins

Number of bins (at least 2).

labels

Bin labels, lowest to highest; default `Low`/`Med`/`High` for 3 bins, `Q1`... otherwise.

missing_label

Label of the missing level; no such level is added if nothing is missing.

reference

Label of the reference level, one of `labels`.

subset_mask

Optional logical vector (length of `x`): compute the quantile breaks on these rows only (for example events only, or the training rows); all rows are then binned with those breaks.

Value

A factor. The breaks are in the attribute `breaks` (inner cutpoints).

See also

[tbl_cox_binned_missing()]

Examples

x <- c(1:9, NA, NA)
table(cut_with_missing_category(x))
#> 
#>     Low     Med    High Missing 
#>       3       3       3       2