I would like to conduct a meta-analysis to understand how an intervention A affects a desired outcome. While collecting data from the studies, I have noticed that some look at the direct effect of A on the dependent variable, while others look at the effects of the interaction of A with other factors (which are unique for each study) on the DV. My question is whether it is possible to utilise all these effect sizes in my meta-analysis or would it confound my results if I combined main and interaction effects?


1 Answer 1


The main effect for a variable in the presence of an interaction that includes that variable has a different meaning and interpretaion than in a model where there is no interaction.

In the presence of an interaction, the main effect is conditional on the variable that it is interacted with being zero (or at it's reference level in the case of a categorical variable). This is very different from a model without interactions where the estimate can be interpreted as a partial regression coefficient.

So coefficients for main effects are not comparable and you shouldn't include them both in a meta analysis

  • $\begingroup$ Thank you for your reply! In this case, would it be sensible to aggregate the effect sizes from the studies that look at the direct effect, as well as the main effect of treatment A from the studies that involve an interaction with another factor? Alternatively, would it make sense to split the dataset into direct effects and interaction effects and perform separate meta-analyses? Thank you! $\endgroup$
    – PK89
    Mar 1, 2021 at 14:28

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.