Tools for Those Who Summarize the Evidence Base
I am looking for a software package - preferably in R - that allows me to use user-specified weights in meta-analysis (instead of inverse-variance weights).
I am looking at a collection of randomized controlled trials. The issue is that I have some trials that are cluster-randomized, whereas some are individual-randomized. I want to "down-weight" the individual-randomized trials, and I'm not sure how to do that. I know that standard practice is to use inverse-variance weighting and use ANOVA or meta-regression to deal with the issue, but I want to explore this alternative specification.
Let me know if you can help!
Aah, difficult problem.
One option would be winsorizing the individual-randomized trials so that they do not exceed the weight given to the cluster-randomized trials and then re-analyzing. And you might apply some other rule for doing that (more extreme, less extreme). But of course you will want to evaluate and document whether different versions of winsorizing change your inferences.