The Merits of Using Longitudinal Mediation

Abstract

Many of the mediation analyses reported in the literature are based on concurrent or single-occasion data sets. The 2 overarching themes of the present article are: Results of concurrent mediations are inherently ambiguous, and researchers would be wise to conduct mediations on longitudinal data sets instead. An example included here demonstrates that a significant concurrent mediation result is not necessarily supported by a longitudinal mediation analysis conducted with longitudinal data. Examples are presented of 4 basic longitudinal mediation types: (a) explaining a mean group difference over time, (b) a mediation analysis embedded in an experimental design, (c) a focused longitudinal mediation (1 specific ordering of variables over time), (d) a complete longitudinal mediation (all possible orderings of variables over time). Obtaining longitudinal data is often resource intensive, but if researchers want to obtain clear and unambiguous mediation results, it is advised that they perform these analyses on such data sets.