[R-meta] robust variance estimator in meta-analyses of rare events (proportions)

[Viechtbauer, Wolfgang (SP)]

Hi Pier-Alexandre,

If you have results in terms of log odds (which is what a GLMM approach will give you), then you can always back-transform estimates and corresponding CIs to proportions using plogis().

But as Guido said, I see no immediate need for using a robust variance estimator in that case (unless you have other forms of clustering, but then you might as well try to model them directly via the GLMM).

Best,
Wolfgang

-----Original Message-----
From: R-sig-meta-analysis [mailto:r-sig-meta-analysis-bounces using r-project.org] On Behalf Of Guido Schwarzer
Sent: Thursday, 04 October, 2018 17:19
To: r-sig-meta-analysis using r-project.org
Subject: Re: [R-meta] robust variance estimator in meta-analyses of rare events (proportions)

Am 04.10.18 um 15:50 schrieb Pier-Alexandre Tardif:
> Dear Guido,
>
> Yes I already tried this method (I replicated http://www.metafor-project.org/doku.php/analyses:stijnen2010), results were slightly different from that obtained using the previous classic code but it's hard to tell which is more or less biased given the many differences involved!?

For rare binary data, I would postulate that GLMMs taking the binary 
structure into account are more appropriate than classic methods 
assuming normality of the transformed proportions within individual studies.

> Moreover, for a meta-analysis of proportions we can only use the logit transformation with rma.glmm (double arcsine is not available) and I still wouldn't know how to get the back-transformed values if I use a robust variance estimator (e.g. with clubSandwich package). Any ideas from here?
I see no problem in using the logit transformation. Futhermore, I do not 
think that using a robust variance estimator is necessary if you are 
using GLMMs. Maybe somebody else could comment on this.

Best wishes,
Guido