[R-meta] Influential case diagnostics in a multivariate multilevel meta-analysis in metafor

[Yogev Kivity]

Hi all,

I am fitting a multivariate multilevel meta-analysis in metafor and having
trouble computing outlier and influential case diagnostics (i.e., cook’s
distances per
https://wviechtb.github.io/metafor/reference/influence.rma.mv.html).

This a large dataset of 3360 Pearson’s correlations (converted to Fisher’s
z) nested within 600 subsamples that are nested within 311 studies. Below
is the code I used for the model and for computing Cook’s distances, and
the problem is that it takes it a lot of time to run (I ran it overnight
and it only reached 6%). I am assuming it is related to the size of the
dataset and to the complex model structure, but I am not sure how to go
about and speed up the processing. I should note that I am computing the
distances based on the simplest possible model (i.e., no moderators and
without considering dependencies among effect sizes within clusters).

I was hoping someone could help with some suggestions of how best to move
forward.

Thanks,
Yogev

NoMods <- rma.mv(yi, vi, random = ~ 1 | StudyID/GroupID/EffectSizeID,
data=Data,sparse=TRUE)
summary(NoMods)
NoModsCooksDistance <- cooks.distance(NoMods,progbar = T,cluster = StudyID,
reestimate=FALSE,parallel="snow")
NoModsCooksDistance
plot(NoModsCooksDistance, type="o", pch=19)


--

Yogev Kivity, Ph.D.
Postdoctoral Fellow
Department of Psychology
The Pennsylvania State University
Bruce V. Moore Building
University Park, PA 16802
Office Phone: (814) 867-2330

	[[alternative HTML version deleted]]