I'm running a meta-regression and am inputting at different study characteristics as IV's. The problem is some are continuous and some are categorical. The IV's are: Age Gender (2 values, dummy coded) Study design (2 values, dummy coded) Population type (4 values, dummy coded).

I did univariate analyses first, and age and study design were significant. I also did a univariate analysis with population type, and one of them was significant (although I understand this means it is only significant from the reference category).

When I put all the significant ones into a multiple meta-regression (i.e. age, study design, populationtype1, populationtype2, populationtype3), only one of the population type dummy codes is significant.

My questions are: Can you put two (or more) categorical iv's in one analysis (does the constant apply to both categorical variables e.g. the reference study design and population type)? How do I interpret this if this is the case? What further analysis should I do to unpick the effect of population type?

I hope this makes sense


1 Answer 1


Since meta-regression is very similar to standard regression you can do anything you could do in the standard form in meta-regression. So it is fine to include as many moderators as you wish of whatever type, categorical or continuous. Just as in standard regression though the results you get will depend on the correlation between the moderators and the number of primary studies you have.In your case you have six moderators (counting population type as three) so I would have thought the minimum number of primary studies to get a sensible result was between 50 and 100.


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.