[R-meta] Results interpretation publication bias
||@t@ @end|ng |rom dewey@myzen@co@uk
Thu Apr 6 15:05:27 CEST 2023
The visual impression of your funnel plot is affected by having a few
points corresponding to small studies. Can you get a better picture if
you eliminate them? Since, from memory, most of your studies have sample
sizes in the hundreds a few smaller ones are not really going to affect
any scientific conclusion.
On 06/04/2023 09:49, Dr. Guido Schwarzer via R-sig-meta-analysis wrote:
> Dear Wilma,
> Despite having a significant Thompson-Sharp test for funnel plot asymmetry (BTW, I think you didn't report the p-values for the Egger and Begg-Mazumdar tests), the result of the limit meta-analysis basically says that there is no real problem with small study effects as the bias-adjusted and the original random effects estimate are very close to each other: 0.39 [0.33; 0.44] vs 0.36 [0.31; 0.41]
> Would this difference change your overall conclusions at all?
> Concerning the one potentially influential study, you could run a sensitivity analysis by excluding the study and running the Thompson-Sharp test and other analyses on the reduced data set.
> [[alternative HTML version deleted]]
> R-sig-meta-analysis mailing list @ R-sig-meta-analysis using r-project.org
> To manage your subscription to this mailing list, go to:
More information about the R-sig-meta-analysis