Modeling longitudinal mediation of within-person changes

Mediation analysis concerns the discovery of mechanisms that transmit an effect from one variable to another. At times, social scientists propose hypotheses involving mediation of within-person changes (e.g., do within-person changes in cognitive ability mediate the relation between changes in depression and changes in disability?) The current manuscript introduces an approach within the latent change score (LCS) modeling framework to test such hypotheses. This approach is compared to existing methods, including the cross-lagged panel model (CLPM; Cole & Maxwell, 2003) and the LCS mediation model (Selig & Preacher, 2009). Results from (1) a single-replication simulation and (2) an empirical example utilizing data from the Health and Retirement study are presented.