Facilitating Model Comparison and Evaluating the Impact of Model Specification Using Simulation Studies With MplusAutomation

• In Event: Machine Learning Algorithms, New Modeling Approaches, and Simulation Methods to Understand Model Performance

Abstract

Over the past 20 years, Monte Carlo simulation studies have helped establish modeling practices in the larger latent variable framework. Impactful studies like Hu & Bentler (1999) provide model fit recommendations that shape how researchers use models in practice. Additionally, simulation studies provide a range of other important modeling considerations-- including studying the needed sample size for sufficient power, the robustness of a particular estimator, the impact of violating model assumptions, and understanding which fit indices are most helpful during class enumeration in mixture modeling.
In a definitive study, Muthen and Muthen (2002) provide a useful introduction to using Monte Carlo simulations within MPlus to help researchers determine the appropriate sample size while maintaining high power. To further assist researchers, we expand upon Muthen and Muthen's study to include a greater range of models with additional conditions and use MplusAutomation for Monte Carlo simulations. MplusAutomation, a package in R, significantly speeds up the specification of simulation population models, provides validity checks throughout the simulation process, helps reduce user errors, and quickly organizes the results. The models used in this demonstration include a regression model with dichotomous indicators, a two-factor model, and a growth model. Conditions, such as missing data, normality, and sample size (e.g., 100 through 2000), were varied to determine their impact on power, bias, and coverage. We will also provide context for using MplusAutomation with simulation studies for mixture models. This demonstration intends to encourage researchers to use and enhance MplusAutomations' capabilities for Monte Carlo simulations.

Authors

• Netasha Pizano, University of California - Santa Barbara

• Dina Naji Arch, University of California - Santa Barbara

• Delwin Carter, University of California - Santa Barbara