Abstract

Research is increasingly moving toward intensive longitudinal methods to look at dynamics in psychological constructs such as affect and well-being. Still, researchers typically assume that the measurement model (MM) — the way items relate to underlying latent constructs — is not changing across time (i.e., longitudinal measurement invariance holds) or see these changes as nuisance. Not only is this often not realistic, but it completely ignores the valuable insights that can be gained from embracing and teasing apart the dynamics of the MM itself and how it relates to characteristics of the individuals (e.g., personality) and the contexts (e.g., onset of a stressful event). Studying MM dynamics can uncover context-specific changes in, for example, emotional granularity, engagement, and affect experience, and ultimately contribute to theory building. The latter is crucial in emerging fields like intensive longitudinal research that are still light on theory. A method ideally suited for uncovering these dynamics is Latent Markov factor analysis (LMFA; Vogelsmeier et al., 2019), 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). In this presentation, I will describe how LMFA reveals MM differences across individuals and time as well as possible reasons for these dynamics, and illustrate LMFA with empirical applications. I will also introduce the user-friendly software package “lmfa” that allows researchers to easily embrace MM dynamics in their own intensive longitudinal data.