Studying the Strength of Prediction Using Indirect Mixture Modeling: Nonlinear Latent Regression with Heteroskedastic Residuals

A framework of R-squared measures for single-level and multilevel regression mixture models.

“…As a second clarification, our R 2 measures provide meaningful information about a regression mixture model even when prediction is not the primary objective of the analysis. Though mixture models are increasingly being used for individual prediction (e.g., Cole & Bauer, 2016; Cudeck & Henly, 2003; de Kort, Dolan, Lubke, & Molenaar, in press; Sterba & Bauer, 2014) it is still more common for analysts’ primary objective to involve description and/or explanation of theorized data-generating processes. In the latter context, our R 2 ’s serve as relatable, quantitative effect size measures that aid substantive interpretation, which has heretofore usually been qualitative and informal in mixture contexts.…”