CODEWITHSUNDEEP
rlme4metafor
HJT
HJT

Reputation: 327

Multilevel Meta-Analysis Using metafor - predictors at multiple levels

I am running a meta-analysis of correlation coefficients within papers. So lvl 1 = papers and lvl 2 = coefficients within papers.

I'd like to include moderators at the paper and the coefficient level, but I'm unsure how to specify the level at which the moderator should be applied. For example, how do different methods at the paper level effect the outcome and how do different participants at the coefficient level effect the outcome.

My current model is:

m1 <- rma.mv(yi, vi, mods = ~ paper_factor + coefficient_factor, random = ~1|paper/id, data=dat)

But, this doesn't specifically encode that the paper_factor differs between papers and the coefficient_factor differs between coefficients within a paper. How would I do that?

Note: metafor syntax is similar to lme4 so perhaps the appropriate syntax there would work here as well?

Upvotes: 1

Views: 224

Answers (1)

Wolfgang
Wolfgang

Reputation: 3395

The 'level' of a moderator is determined by how it is coded. If it is a constant within papers, then it is a moderator at the paper level. If the moderator can vary within papers, then it is a moderator at the estimate level. And indeed, that is also how things work with lme4.

Upvotes: 1

Related Questions