[R-meta] Meta vs Metafor - trapped in a loop

Joao Afonso jot@|on@o @end|ng |rom gm@||@com
Tue Feb 18 16:25:21 CET 2020

Dear all,

I am sorry to come to you once again with my questions about the
meta-analysis results. I thank all the help provided so far but sadly
it seems that can't get robust result. I will try to explain the
problem as thoroughly as possible. My meta-analysis follows the steps

1. meta-analysis with the whole set of studies
2. outliers and influential case diagnostic
3. removal of studies identified in point 2. and re-run the meta-analysis
4. sub-group analysis and meta-regression

As explained before I am having mismatching results when comparing the
output produced by the code I am using when running the meta-analysis
(metafor) and the code I am using when producing the forest plot

As explained by Guido these packages use different harmonic means to
back-transform the proportions. I thought the problem could be there
and thus tried to be consistent using the meta package to run the
meta-analysis. Unfortunately I couldn't do the outliers and
influential case diagnostic as "no applicable method for 'influence'
applied to an object of class "c('metaprop', 'meta')"".

Is there a way to convert a metaprop object into an object that can be
used in a influential analysis? Any other routes I should consider for
solving this problem?

I am desperate for help.

Wishing all a great afternoon,


João Afonso
DVM, MSc Veterinary Epidemiology
PhD Student
Department of Infection and Global Health
University of Liverpool

More information about the R-sig-meta-analysis mailing list