Hi Greg,

I have two potential answers for you, depending on whether you have data on effects between your 6 IVs and some outcome; or data on one or more relationships defined by some effect size, for which your 6 IVs will subsequently act as moderators of that relationship. A key differentiator is whether you are doing the MA also without those IVs you mention, if you can, the second answer makes more sense.

If the first, you need to specify why it matters to include them in one analysis. Could you not simply run 6 seperate MAs? Or one MA with 6 subgroups? What defines the connection between them?

For the second, Meta-regression is indeed what you are looking for. The specific software you refer to may not allow for testing multiple predictors simultaneously, but others do. Specifically R is able to do everything what you want, using packages as metafor (my preference) or meta.

Best,

Robert

Hi Robert Suurmond,

Thank you for your quick and thorough answer!

The IVs relate to the individual protocols of each original study. After operationalizing these protocols I have 7 dichotomous variables that seem to vary between protocol.

I am absolutely open to doing a meta-regression, would it work for multiple dichotomous variables?

Thanks for your insights,

Greg

Hi Gregory,

Yes, meta-regression works with dichotomous and continuous variables (and simultaneously). The estimates of the influence of a predictor on the effect show you the increase or decrease in effect when the predictor increases by 1, compared to the baseline. For dichotomous variables (0/1), this means the change in effect when the predictor is 1, rather than 0. For continuous variables, it depends a bit on the scale, but works in a highly similar manner.