Hi there! I have a quick question: A friend is working on a moderated mediation path model with 3 measurement points (interval scaled). The model basically looks like Xt1 --> M't2 --> M''t2 --> Xt3, so we have baseline control for the outcome. I suggested that we should probably also control for the t1 baseline measure of M' and M'', however, I am also not quite sure and cannot find any paper on this issue right now. Any suggestions would be very much appreciated!
Hi Julian,
I describe this issue (in a general context) in the attached paper. It is important to control for the baseline level of the mediator (and probably the outcome as well) for meeting the assumption of sequential ignorability. If not, depending on the initial levels of these variables, your data may be biased due to regression to the mean.
I hope this helps
Ariel
Article Using mediation analysis to identify causal mechanisms in di...
the best way to assess variables mediation role is Baron and Kenny method.
relying on the recommended stages by (Baron and Kenny, 1986) the following stages;
1) Examining the direct impact of primary exogenous
variables on endogenous variables
2) Examining the impact of exogenous variables on
mediator variables
3) Examining the impact of mediators, on endogenous by control of exogenous variables
4) Study of the regression coefficients’ decrease from stage
1 to 3:
Hi Alireza,
thank you very much for your reply. I would recommend reading "Reconsidering Baron and Kenny: Myths and Truths about Mediation Analysis" by Zhao et al. from 2010. I think you refer to the "classic" mediation-testing?
Kind regards
Julian
- Badges
- Science topic