Check multivariate dispersion differences across metadata variables

check_dispersion() runs vegan::betadisper() and permutation tests on a precomputed sample distance matrix. Strong dispersion differences indicate that PERMANOVA effects may partly reflect spread differences rather than centroid shifts.

Usage

check_dispersion(distance, metadata, variables, n_perm = 999)

Arguments

distance

A dist object containing sample-to-sample distances.

metadata

A data frame with sample metadata.

variables

A character vector naming metadata variables to test.

n_perm

A single positive integer giving the number of permutations.

Value

A data frame with one row per variable.

Examples

data("toy_moat")
metadata <- as.data.frame(SummarizedExperiment::colData(toy_moat))
distance <- compute_biome_distance(toy_moat, distance = "bray")
check_dispersion(distance, metadata, variables = c("outcome", "batch"), n_perm = 99)
#>   variable    status n_groups    statistic p_value risk error
#> 1  outcome evaluated        2 0.0009879928    0.87  low  <NA>
#> 2    batch evaluated        2 0.4410051761    0.50  low  <NA>