Hi All,

I'm trying to determine if it's possible to conduct mediation analysis on three-level data.

I have data with repeated measures at five time points. The predictor variable is at level 3, followed by individuals at level 2, and time point/measurements at level 1. This is based on Rabe-Hesketh & Skrondal's excellent book, and was conducted using Stata. The dependent variable was at level 1, and the predictor variable at level 3.

What I'd now like to do is determine the amount of effect which is mediated through a variable (repeatedly measured at level 1). I can demonstrate that this variable is associated with both X and Y in independent models, but I'm not sure how to proceed with a mediation analyses. All of the literature and examples I've seen deal with two level models, and don't seem to address time.

What options are available?