[R-meta] error for meta-regression with continuous moderator

[Philippe Tadger]

Hello David

I think the main problem is not too many observations additionally with 
a "rare events" (the observation 54/54).  My guess is that the 
observation 54/54 is the observation that the actual model is not able 
to fit. In fact, when you use this modified data (53/54 instead 54/54):

event<-c(28,50,59,29,32,100,36,29,9,50,53,43,49,24)
total<-c(55,60,65,43,41,107,40,34,10,58,54,50,60,28)
The meta-regression converge without warnings.

My suggestion is to use a Bayesian model instead metaprop; that aloud 
you to model this "rare event" of 100% of prevalence.

If you want to stick with metaprop, you have two options:

1) Use method="Inverse" (instead of "glmm"). With this approach you can 
potentially produce bias in the estimation, but don't have any 
convergence issue.
2) Avoid to use moderator as continuous (like using 2 or 3 subgroups)



On 29/10/2021 00:13, david detullio wrote:
> Hello all, I am performing a mixed-effects meta-regression of proportions
> with a continuous moderator. I receive the error message: "Ratio of largest
> to smallest sampling variance extremely large. May not be able to obtain
> stable results."
>
> I saw a post asking about this error, but the post appeared to be referring
> to subgroups, as opposed to a continuous moderator. So, is it common for
> this to happen with a continuous moderator, especially if there are some
> values for the continuous moderator which consist of only a single study?
>
> I have provided the data and code below to produce the results I found. Any
> help would be appreciated.
>
>
> event<-c(28,50,59,29,32,100,36,29,9,50,54,43,49,24)
> total<-c(55,60,65,43,41,107,40,34,10,58,54,50,60,28)
> moderator<-c(5,5,1,11,9,9,5,9,12,2,3,7,5,9)
> data<-data.frame(event,total,moderator)
> meta<-metaprop(event=data$event, n = data$total, sm = "PLOGIT", backtransf
> = TRUE, byvar = data$moderator)
> reg<-metareg(x = meta, ~data$moderator)
> print(reg)
>
> 	[[alternative HTML version deleted]]
>
> _______________________________________________
> R-sig-meta-analysis mailing list
> R-sig-meta-analysis using r-project.org
> https://stat.ethz.ch/mailman/listinfo/r-sig-meta-analysis
-- 
Kind regards/Saludos cordiales
*Philippe Tadger*
ORCID <https://orcid.org/0000-0002-1453-4105>, Reseach Gate 
<https://www.researchgate.net/profile/Philippe-Tadger>
Phone/WhatsApp: +32498774742

	[[alternative HTML version deleted]]