How Within-Person Effects Shape Between-Person Differences: A Multilevel Structural Equation Modeling Perspective

Various theoretical accounts suggest that within-person effects relating to everyday experiences (assessed, e.g., via experience sampling studies or daily diary studies) are a central element for understanding between-person differences in future outcomes. In this regard, it is often assumed that the within-person effect of a time-varying predictor X on a time-varying mediator M contributes to the long-term development in an outcome variable Y. In the present work, we demonstrate that traditional multilevel mediation approaches fall short in capturing the proposed mechanism, however. We suggest that a model in which between-person differences in the strength of within-person effects predict the outcome Y mediated via mean levels in M more adequately aligns with the presumed theoretical account that within-person effects shape between-person differences. Using simulated data, we show that the central parameters of this multilevel structural equation model can be recovered well in most of the investigated scenarios. Our approach has important implications for whether or not to control for mean levels in models with within-person effects as predictors. We illustrate the model using empirical data targeting the question if the within-person association of occurrence of daily stressors (X) with daily experiences of negative affect (M) longitudinally predicts between-person difference in change in depressive symptoms (Y). Implications for other multilevel designs and intervention studies are discussed.