Skip to contents

Audit Rank Deficiency of a Model's Fixed-Effects Matrix

Source: R/inspect_rank_deficiency.R
inspect_rank_deficiency.Rd

With sparse or unobserved factor cells, `lme4::lmer()` silently drops redundant columns of the fixed-effects model matrix, and `lm()` / `glm()` return `NA` coefficients. This reports which terms were dropped.

Usage

inspect_rank_deficiency(model, verbose = TRUE)

Arguments

model

A `merMod` object (`lmer`, `glmer`) or an `lm` / `glm` object.

verbose

Emit a markdown callout with [message()] when the model is rank-deficient.

Value

A list of class `cbe_rank_deficiency` with `rank_deficient` (logical), `n_columns` (columns of the full model matrix), `rank`, `dropped` (names of the dropped terms) and `message` (the markdown callout text; `NULL` if full rank), returned invisibly.

See also

[lmer_safely()]

Examples

d <- data.frame(y = c(1, 2, 3, 5, 4, 6), a = 1:6, b = 2 * (1:6))
inspect_rank_deficiency(lm(y ~ a + b, d))$dropped
#> > **Rank-deficient model matrix.** 1 of 3 fixed-effect column(s) were dropped (rank 2): `b`. The cells that define them are empty or confounded with other terms, so their effects are not estimated.
#> [1] "b"