What are the core assumptions when doing mediation analysis in structural equation modelling? I am estimating the classic mediation model using SEM in panel data in Stata with fixed effects (image attached). Having found a relationship between X and Y in my earlier fixed effects regressions, I want to look at the importance of mediator variables in my data in explaining this relationship, i.e. does information on net worth mediate the relationship I observed between unemployment shocks and mental health scores. I know how to do this in my statistical program, but first I would like to know what the core assumptions (implicit or explicit) are in this kind of an analysis.

enter image description here

  • $\begingroup$ Such analyses usually need to assume no unmeasured confounding between mediator and outcome (dependent variable). You also require the usual exchangeability assumption for analyses not involving mediator. $\endgroup$ – hehe Apr 3 at 19:47
  • $\begingroup$ Thanks for your response, could you please explain what you mean by the usual exchangeability assumption for analyses not involving the mediator? $\endgroup$ – John Apr 18 at 15:38

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.