CMStatistics 2021: Start Registration
View Submission - CMStatistics
Title: Latent Markov factor analysis for evaluating measurement model heterogeneity in intensive longitudinal data Authors:  Jeroen Vermunt - Tilburg University (Netherlands)
Kim De Roover - Tilburg University (Netherlands)
Leonie Vogelsmeier - Tilburg University (Netherlands) [presenting]
Abstract: When studying intensive longitudinal data (e.g., with Experience Sampling Methodology), drawing conclusions about dynamics of psychological constructs (e.g., well-being) over time requires the measurement model (MM; indicating which items measure which constructs) to be invariant between subjects and within subjects over time. However, there might be heterogeneity or non-invariance in the MM, for instance, due to subject-specific differences and changes in item interpretation or response styles. Mixture modeling approaches have proved to be powerful tools to detect unobserved heterogeneity, but the methodology to evaluate measurement invariance for multiple time points and subjects simultaneously was lacking. To fill this gap, we built upon common mixture modeling approaches and proposed latent Markov factor analysis (LMFA), which combines a discrete- or continuous-time latent Markov model (that clusters observations into separate states, according to state-specific MMs) with mixture factor analysis (that evaluates which MM applies for each state). We introduce this novel methodology, illustrate it by means of an empirical example, discuss two possible estimation procedures, and explain the latest extension, latent Markov latent trait analysis (LMLTA), that adequately deals with ordinal responses.