I ran a logistic regression, all assumptions checked and so on, and I have two models with four predictory variables predicting one dependent. In my first model I just check the variables themselves, and only variable 2 is significant. In my second model I include all possible interaction terms between my four predictors and now one of these interactions becomes significant.

For my conclusion, do I just show the odds ratios of the models separately? And how do I explain the significant interactions in words? I understand that for variables I can say: This variable has a significant effect and its odds ratio is Exp(B) so we can see that category 1 occurs 3 times as probable as category 2 for instance. But how do I say this for interaction terms that are significant with their odds ratios?

Thanks if anyone could help me and happy holidays guys!


1 Answer 1


Follow up question: If your interactive term is significant, how did you assess this? If you are just using the P-value, you may wish to use a likelihood ratio test to determine which model is a better fit.

If the likelihood ratio test reveals that your interactive model is a better fit, you would report the odds ratios for each level of the interactive term in question. You may wish to report the odds ratio for the non-interaction model, but this will depend on your research question. Generally, you should only be concerned with the interaction, not the crude value without interaction, as by definition the crude value would be misleading.

So to summarize, report the odds ratios at each level of the interaction. Additionally, the odds ratio will still be compared to your reference group, not to other positively coded values in the interaction, so be careful in saying this category occurs x times more than this category. Hope this helps.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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