Latent profile analysis (LPA) is an emerging approach to analyze the Revised Illness Perception Questionnaire (IPQ-R). LPA creates subgroups with similar illness perceptions. We used simulated data sets to provide suggestions and considerations for IPQ-R researchers implementing LPA. We explored 640 simulation parameters, varying sample size, IPQ-R distribution, covariance, and subscale means, simulating 3 distinct latent subgroups. We simulated 1000 samples for each setting via MClust package in R. Caution should be used when N N ⩾ 100 still may not yield ideal performance depending on sample (e.g., subgroup sizes, within-group variance). With more differences between subgroups, LPA is more accurate. However, researchers have little control over mean differences, except indirectly (e.g., diverse sample). Researchers using LPA with IPQ-R data must carefully consider anticipated sample heterogeneity to establish appropriate sample size estimates. Resources provided in this manuscript can support these determinations.
Westra et al. (Sun,) studied this question.