I am currently conducting a moderated mediation analysis using Model 7 in the PROCESS macro. I want to control for demographic variables such as age, organizational sector, and tenure.
Should I run two versions of the model — one without the covariates and one with them — to assess whether these demographics meaningfully influence the model? When I include them, the output shows their effects only on the direct paths, but not on the indirect or conditional effects. Is this output sufficient, or is it necessary to compare models with and without covariates to assess their impact on the overall model?
What you suggest (running your analyses with and without covariates) makes sense to me. That way, you can check whether inclusion of the covariates changes the direct effects among the variables that are of substantive interest to you.
If you have additional specific hypotheses about the covariates exerting interactive (in addition to additive) effects, you could test this by including product terms (interactions) involving the covariates in yet another model.
Thank you! Is this common practice or do researchers usually just run one model with covariates in ?
my advisor told me to control for them in the anlaysis so not sure if i am doing to much by having one with and without covariates
- Badges
- Science topic
- Similar topics
- Geoscience
- Asia
- India