I'm looking at question #1 on the homework and I'm not sure exactly what type(s) of regression(s) to run. According to the PowerPoint slide handout for "Additive, Moderated, and Mediated Models", Slide 7, we are supposed to run three separate regressions:
1. Outcome by Predictor
2. Outcome by Moderator
3. Outcome by Moderator/Predictor Interaction Term
Did I understand this right? Also, are we running three separate linear regressions, or one multiple regression, or some combination of linear and multiple regressions?
Subscribe to:
Post Comments (Atom)
2 comments:
I do not have the homework in front of me, so I cannot tell you if you have chosen the correct regression to run for #1. However, I think I can answer the second part of your question. I believe you are looking at the third major bullet point on slide 7. If so, you enter the predictor, moderator, and the interaction term all at once.
If you are testing moderation using a calculated interaction term, you must run one and only one regression. The regression must have the interaction term entered in the last step and the predictor & moderator entered on one or more previous steps (usually one, less commonly entered in separate steps, because that's really not the question you are answering).
If you are testing moderation with categorical moderators, you must run the number of regressions that there are levels of the moderator variable, and the only variables in the equations are the predictor(s) and the outcome. (That is, the moderator is used to split the file, so it cannot be a variable in the regression, as it would not vary in the split files--all individuals in group 1 would have a value of 1 for the moderator, all in group 2 would have 2, etc.)
Post a Comment