Sample size requirements with unbalanced subgroups in latent growth models

Multiple group latent growth models enable these comparisons. However, the combination of a smaller subgroup with a larger reference group has been shown to cause issues with power and Type I errors.

The current study explores the limits of the subsample sizes in latent growth modeling (LGM) that can and cannot be analyzed with Maximum Likelihood and Bayesian estimation, where Bayesian estimation was examined not only with uninformed, but also with informed priors. The results show that Bayesian estimation resolves computational issues that occur with ML estimation, and that the addition of prior information can be the key to achieving sufficient power to detect a small growth difference between groups.