Permutation test for a binary covariate's effect on topics

Refits the model many times with the treatment labels permuted, aligning topics across refits, to build a null distribution for the treatment effect on each topic (cf. stm::permutationTest). Fast because each refit is cheap.

Usage

permutation_test(
  formula,
  model,
  treatment,
  corpus,
  nruns = 100L,
  seed = NULL,
  ...
)

Arguments

formula: Prevalence formula whose RHS includes treatment.
model: A faSTM fit.
treatment: Name of a 0/1 covariate in corpus$meta.
corpus: The faSTM_corpus the model was fit on.
nruns: Total models (1 reference + nruns-1 permutations).
seed: RNG seed.
...: Passed to stm() for the refits.

Value

A faSTM_permtest with ref (observed per-topic effects) and null ((nruns-1) × K permuted effects).