Latent transition analysis (LTA) is a mixture modeling approach that is gaining popularity in social science, behavioral, and health research. LTA is a longitudinal method that can be used to investigate how individuals transition from one latent, or unobserved class, to another over time. Although LTA is gaining use in many disciplines, to date only two studies have examined the statistical power of this statistical approach. The present study aims to examine how sample size and model characteristics such as latent transition probabilities, model definition, item-response probabilities, and class size influence the statistical power of to detect effects in latent transition probabilities. Meta-analysis findings were used to guide conditions ultimately used in this Monte Carlo simulation study. All data were generated using Mplus (Muthén & Muthén, 1998-2014).

Results from this study revealed how larger sample sizes, larger transition probabilities and class sizes were more likely to have greater power. Results also highlighted the importance of a well-defined measurement model with high class separation and homogeneous classes and its influence on statistical power. Findings from this dissertation provide evidence on which conditions tend to have higher or lower power. Additionally, findings show how poor conditions can have model convergence issues and provide misleading results due to "artificially high" power values. This study also includes practical recommendations and suggestions for future directions.