I'm running a survey where I'd like to compare three models with the same set of variables except for the moderator. I'm looking for the strongest interaction effects as a foundation for subsequent experiments.
To be more specific, I would run hierarchical regression analyses with the same covariates, and the same IV, but a different moderator and a different interaction term for each model. All measured on the same sample: Step 1 The main IV ; Step 2 The moderator and the interaction term ; Step 3 the covariates.
Then I would do a regression analysis with all the predictors plus the interaction terms, and a commonality analysis, to see which predictor/ interaction explains more variance compared to others.
Assuming assumptions are met, assuming no multicollinearity, assuming interaction terms are centered ... I was wondering: does it makes sense? Do I have to plan some comparisons in advance to not to lose power? If so, how to do that? Doesn't make sense? If so, why?
