blrt: Bootstrap likelihood ratio test (non-regression only)

In poLCAParallel: Polytomous Variable Latent Class Analysis Parallel

Bootstrap likelihood ratio test (non-regression only)

Description

Does the bootstrap likelihood ratio test. Provide two fitted models, the null model and the alt model fitted with a different number of clusters/latent classes. Bootstrap samples are generated using the null model. The null model and the alt model are refitted on the bootstrap samples to investigate the log likelihood ratio of the two models.

Usage

blrt(
  model_null,
  model_alt,
  n_bootstrap,
  n_thread = parallel::detectCores(),
  n_rep = 1,
  max_iter = 1000,
  tol = 1e-10
)

Arguments

model_null	Fitted poLCA object, the null model
model_alt	Fitted poLCA object, the alt model
n_bootstrap	Number of bootstrap samples
n_thread	Number of threads
n_rep	Number of initial values to try when fitting on the bootstrap samples
max_iter	Maximum number of iterations for EM algorithm
tol	Tolerance for difference in log likelihood, used for stopping condition

Details

Runs in parallel for each bootstrap sample, potentially high memory if the data is large

Value

List containing the following:

• fitted_log_ratio: log likelihood ratio comparing the null and alt models

• bootstrap_log_ratio: vector of length n_bootstrap, bootstrapped log likelihood ratio comparing the null and alt models

• p_value: the proportion of bootstrap samples with log likelihood ratios greater than the fitted log likelihood ratio

poLCAParallel documentation built on Feb. 20, 2026, 1:09 a.m.