Thinking within-persons: Using unit fixed-effects models to describe causal mechanisms

This is the authors' abstract. We don't add key points for this paper.

Psychological research is mainly concerned with causal questions, and as a consequence can benefit from explicitly adopting a causal inference framework. In this paper, I explain that the main contribution of longitudinal analysis from a causal perspective is the ability to control for time-invariant unobserved heterogeneity, which can be achieved by focusing exclusively on within-person variation. Based on this general framework, I describe different procedures for estimating within-person effects, as well as modelling strategies that can be used to test substantive hypotheses regarding within-person asymmetrical causation, moderation, effect heterogeneity, and reciprocal causation. These statistical techniques can expand the methodological tool kit for describing mechanisms using observational data. I provide an empirical illustration of these methods by estimating the within-person effects of executive functions on academic achievement.