My thesis includes three independent variables (IVs), one mediating variable (MV), one moderating variable (MO), one dependent variable (DV), and four control variables (CVs). None of these variables are measured as latent constructs.
In the mediation analysis, should I include all the variables — the three IVs, the MV, and the four CVs?
In the moderation analysis, I plan to examine the relationship between the mediating variable and the dependent variable, with the moderating variable as an interaction term. Is it appropriate to include control variables in this moderation model?
Yes, you can include all these variables (and the interaction term(s)) into a single statistical model using path analysis.
Tram Tran Yes, you should include your four control variables in both your mediation and moderation analyses to ensure the robustness of your findings. For the mediation analysis, include all three IVs, the MV, and the four CVs to isolate the unique relationships and prevent omitted variable bias. For the moderation analysis testing the MV*MO interaction on the DV, you must also include the CVs to control for potential confounders, ensuring the observed moderation effect is accurately attributed to your theoretical variables; just remember to mean-center the MV and MO before creating your interaction term to aid interpretation and reduce multicollinearity.
- Badges
- Science topic
- Similar topics
- Computer Science
- Research Topics