I conducted a meta-analysis of proportion on R. After I obtained the funnel plot little asymmetry, I used Egger's regression test and rank correlation test to evaluate the publication bias: for Egger's regression test---> p=0.0322 (presence of publication bias) and rank correlation test ---> p = 0.1635 (absence of publication bias).
In this what should I do?
If you are sure of your computations, Then report your metanalysis. In research we check everything and then call them as we see them. Best, D. Booth
Hello,
I tend to report publication bias when both tests signal bias. In most cases, both are rather inconsistent which does not really strengthens their interpretation.
In a recent review process, the reviewer recommended to me this new paper. As we were in a hurry, and could afford not to comply, I did not read it. Perhaps it is useful for you.
https://link.springer.com/article/10.1007%2Fs10869-020-09687-3
Besides that, I agree with David. A result is a result and concluding publication bias is good basis for a critical part in your discussion section. In addition, you could estimate the trim-fill corrected estimate to inform the field about the severity of the bias.
All the best,
--Holger
You found small study effects, not publication bias. You have to think and give some explanation on why smaller studies show systematically different results than larger studies. Is it due to publication bias or some other reason (e.g. heterogeneity). Either case you report the results.
You should first learn and understand what significance tests are about :/
if there is publication bias on egger test you can conducted fill and trim analysis
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Regression