Observations on the Use of Growth Mixture Models in Psychological Research

Abstract

Psychologists are applying growth mixture models at an increasing rate. This article argues that most of these applications are unlikely to reproduce the underlying taxonic structure of the population. At a more fundamental level, in many cases there is probably no taxonic structure to be found. Latent growth classes then categorically approximate the true continuum of individual differences in change. This approximation, although in some cases potentially useful, can also be problematic. The utility of growth mixture models for psychological science thus remains in doubt. Some ways in which these models might be more profitably used are suggested.Growth mixture models (GMMs) are designed to separate a general population of individuals into subgroups characterized by qualitatively distinct patterns of change over time. In this article, I offer a few observations on the application of these models in psychological science. Like many, I was initially excited about the potential of GMMs. After several years of evaluating these models and reviewing applications, however, I am now skeptical that they will meaningfully advance our understanding of psychosocial development. In what follows, I outline key methodological and theoretical concerns that I have with current applications of GMMs.

A General 3-Step Maximum Likelihood Approach to Estimate the Effects of Multiple Latent Categorical Variables on a Distal Outcome

“…Both methods perform the worst when there is within-class skewness in Z, but the 1-step approach is also likely to alter the number of classes needed to fit the data (consistent with the findings of Bauer, 2007).…”

“…As noted earlier, it is also possible that for the 1-step approach non-normality of Z may lead to an increase in the number of latent classes needed to fit the data (Bauer, 2007). This is a major concern in empirical studies because a change in the number of classes may alter their interpretation and hence bias their estimated effects on the distal outcome.…”

Developmental Links Between Trajectories of Physical Violence, Vandalism, Theft, and Alcohol-Drug Use from Childhood to Adolescence

“…Second, GMMs have been criticized; this method can result in the detection of trajectory classes that are not present in the data (Bauer 2007). One way of minimizing this risk is to not use GMM in an exploratory manner, but rather to specify specific a priori hypotheses based on prior research regarding the shape of the trajectories and the relationship between trajectory membership and covariates (Muthén 2004).…”

Mixture class recovery in GMM under varying degrees of class separation: Frequentist versus Bayesian estimation.

“…Despite the growing popularity and increased usage (see e.g., Bauer, 2007), there are still certain limitations and concerns surrounding GMM that should be mentioned. Specifically, Bauer and Curran (2003) found that multiple latent classes could appear optimal when the population was only comprised of one group and when the data were nonnormal.…”