ref_support

Usage

ref_support(fit, newdata)

Checks each row of newdata against the covariate ranges and the joint (Mahalanobis) spread of the reference data on the covariates the model actually uses. Statuses: "in", "edge" (outside the central 96% of a covariate or beyond the 99% Mahalanobis radius), "out" (outside the observed range or beyond twice that radius), and "new_group" (unseen factor level), and "unknown" (a covariate is NA).

Arguments

Argument	Description
fit	A ref_fit.
newdata	Target data.

Value

A tibble with .row, support, and the squared Mahalanobis distance d2 (NA when there are fewer than two numeric covariates or the row has a missing covariate).

Examples

ref <- ref_simulate(80, seed = 1)
spec <- ref_spec(family = ref_gaussian(), location = ~ age + sex)
fit <- ref_fit(spec, data = ref, outcomes = "y")
ref_support(fit, ref[1:3, ])

# A tibble: 3 × 3
   .row support     d2
  <int> <chr>    <dbl>
1     1 in      0.933
2     2 in      0.325
3     3 in      0.0306