[R-meta] Outlier and influential case diagnostics using generalized linear mixed models approach

[Guido Schwarzer]

Am 17.02.20 um 16:14 schrieb Joao Afonso:

> Dear all,
>
> I am developing a generalized linear mixed model for my meta-analysis
> on lameness prevalence in British Dairy Cattle. Everything seems to be
> working fine however when I try to identify outlier using functions
> rstudent, leave1out and influence R informs that "no applicable method
> ... applied to an object of class "c('metaprop', 'meta')". Is there a
> way to do the diagnostic of influential cases with a rma.glmm object?

The leave-one-out method is available in metainf() of *meta*.

For regression diagnostics provided by *metafor*, you have to conduct a 
meta-regression of your subgroup meta-analysis first.

E.g.,

m = metaprop(nlameanimal, ssizeanimal, author, data=prevalence_2020_noout,
              method="inverse", sm="PLOGIT", method.tau="DL", 
method.ci="NAsm",
              byvar=lcmbi, tau.common=TRUE,prediction = TRUE)

mr = metareg(m)

rstudent(mr)

Best wishes, Guido


	[[alternative HTML version deleted]]