predict function linking metafor:rma.mv with glmulti?

[Kelly Gravuer]

I'm using the metafor package to build meta-analysis models - specifically, the rma.mv() function since I have some random effects.  I also have a few predictor (modifier) variables and am using glmulti to explore their relationships to the effect size (via multi-model averaging as described here, which gives me a table of averaged coefficients).

I would now like to make predictions for a new set of predictor variable values, using the multi-model averaged parameters.  I realize I can calculate the predicted effect size itself using the averaged coefficients directly, but I'd also like to have confidence intervals for the predicted effect size, and I'm not sure of the correct way to get those.  rma.mv() itself has a predict function, so I can easily get predictions + CIs for the best model returned by glmulti, but for the particular question I'm asking, I think multi-model averaged predictions would be better.

The attached vignette suggests that I may need to write a custom predict function that links glmulti and rma.mv.  Before I wade into that, thought I'd check here to see whether anyone has done this already?

Thanks so much for any suggestions!

Kelly