How do you interpret interaction terms between two categorical variables if Y is significant, Z is not significant, but Y*Z is significant. Do you include the interaction term in the model, or do you eliminate the variable that is not significant (in this case Z) and therefore also the interaction term?

  • $\begingroup$ In addition to Ben's answer; making a quick graph of the means of the dependent in all categories for the interaction, Y x Z, might help you with interpreting the result. You can do this on paper no code needed. $\endgroup$ – stijn Apr 30 '15 at 2:24

Keep both the interaction term and the main effects.

Because Z influences the model both through the main Z term and the interaction term Y*Z, it's not very meaningful to look at the significance of the main effect on its own. See for instance this answer:

If the interaction is significant, interpreting either main effect, whether significant or not, is basically pointless (and misleading). The reason is that when $A$ and $B$ are involved in an interaction, the coefficient for $A$ is the effect of $A$ when $B=0$; in other words, the effect is conditional on the value of $B$, and is not a main effect. Similarly, the coefficient for $B$ is the effect of $B$ when $A=0$.

  • $\begingroup$ So making sure I understand you correctly, even though Z is not significant, I should still keep it in the model because it is part of a significant interaction term? $\endgroup$ – stats123 Apr 30 '15 at 2:31
  • $\begingroup$ Yes, that's correct. $\endgroup$ – Ben Kuhn Apr 30 '15 at 5:43
  • $\begingroup$ Since the variables are categorical in this place, replace "$= 0$" in that answer with "equal to the baseline," but it's the same idea. $\endgroup$ – Ben Kuhn Apr 30 '15 at 5:58

Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.