Response to commentaries on target paper, ‘Growth Mixture Modeling in Developmental Psychology’

✍ Scribed by Arin M. Connell; Alice A. Frye

• Publisher: John Wiley and Sons
• Year: 2006
• Tongue: English
• Weight: 99 KB
• Volume: 15
• Category: Article
• ISSN: 1522-7227
• DOI: 10.1002/icd.485

✦ Synopsis

We wish to thank the authors of the three commentaries for their thoughtful responses to our target paper, to highlight some interesting points from the commentaries, and also to provide additional results in response to several suggestions, as they bear on the task of selecting optimal analytic strategies. It may be helpful to clarify that we were asked to submit the target article with the proscribed goal of providing a brief overview of general latent growth mixture modelling techniques to readers of this journal in light of a number of emerging issues regarding the use of mixture modelling techniques. Issues included the proliferation of papers in the developmental literature employing these techniques, the potential density of the topic for researchers not well-grounded in the statistical literature, and the existence of subtle challenges to mixture modelling with which many general readers may be unacquainted.

We first discuss the detailed methodological and conceptual points raised by Muthen. Given space constraints in our original paper, we did not address alternative model choices before proceeding with the mixture modelling analyses. Muthen's commentary thoughtfully underscores the fact that researchers must first consider whether latent growth mixture modelling is the best way to understand the developmental phenomenon under consideration prior to conducting analyses. We bypassed questions about model choice, but we agree that a thorough investigation of alternative models is a critical step in modelling efforts.

We also agree with Muthen on a number of specific analytic points regarding our data set and analyses. We concur with the potential for two-part models to capture change in processes involving skewed responses such as antisocial behaviour, with most youth reporting no engagement in high-risk behaviour at any given time. Indeed, we have used a related technique to analyse substance use trajectories over adolescence in the current sample (see Connell, Dishion, &