Explanation: I have one dependent, one independent, and one moderator. Before I run the PROCESS to check the moderation, I need to check the regression assumptions. Do I need to put the interaction term (independent *moderator) in the box of independent variables and treat the interaction term as a predictor and thus to check the assumptions for the interaction term with the dependent variable (so linear relation interaction term with dependent variable; multicollinearity, homoscedasticity); or do I need just to put the moderator and the independent variable in the box with the independent variables?
Hello Anna,
Yes. Any variable that is part of your model, whether an individual variable or an interaction term, should be included in the model if you want to evaluate the viability of model assumptions for your data set.
Good luck with your work.
@Anna Yes you need to meet all the assumptions of regression including every single variable because moderation analysis is also based on regression.
Anna Christodoulou I agree with the previous posts and just wanted to add an information to make it more clearly why you should include the interaction: if you do not include important interactions (maybe because you do not know that there is an interaction the population) assumptions may fail, e.g. homoscedasticity. In case of missing interaction the residuals may show a funnel which disappears if you include the interaction, because exactly the change of slopes depending on the moderator may account for the funneling, which may disappear then. Therefore, you should always check assumptions (or generally inspect the data) for your fully specified model. If there are still violations, it may be due to missing interactions, higher order polynomials or chosen a wrong distribution of errors (e.g. OLS for count data).
- Badges
- Science topic
- Similar topics
- Biological Science
- Kinesiology
- Running