I carried out hierarchical regression model using stepwise method and included interaction terms (predictor variable x moderator variable) in the third step of the model. It was found that only interaction term successfully entered into the model, but not the relevant predictor and moderator variables. I was wondering what this result means, thank you so much for your help.

  • $\begingroup$ not enough time ATM to write a proper answer, but: when interax are included, main effects parameters refer specifically to the expected effect of a unit change in the parameter at the baseline level of the other parameter. Many stepwise methods don't let you violate marginality by dropping main effects while the interactions are present in the model, precisely to avoid this confusion ... $\endgroup$ – Ben Bolker Jul 25 '18 at 2:38
  • $\begingroup$ Stepwise regression is a bad idea, as there is very little guarantee of selecting the best model given your variables. Other thing is if the interaction makes sense without the main effect, then this may not be a problem. Consider two factor situations, sex and handedness. The interaction simply refers to cross of both factors which is four real life groups, so it's like a one factor with four groups: female with right handedness, male with right handedness, ... Your significant p-value for these four real groups is evidence that not all group means are the same. $\endgroup$ – Heteroskedastic Jim Jul 25 '18 at 11:49

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.