Statistical Power to Detect the Correct Number of Classes in Latent Profile Analysis

Short summary

Latent profile analysis (LPA) often struggles to detect the correct number of classes, especially with small interclass distances. This simulation found that only very large interclass distances (Cohen's d = 1.5) allowed methods like the LMR test, adjusted LMR, bootstrap likelihood ratio test, BIC, and sample-size-adjusted BIC to reliably select the correct number of classes.

AI-generated from the title and abstract; the full text is not read.