Dear Researchers,
I am asking this question for my friend.
"How to comment on/report a pooled proportion effect size calculated through double arcsine transformation method?
Problem statement: We have a pooled estimate of proportion and my observed effect sizes (proportions from included studies) are severely skewed (range: 0.026 - 0.353, while normally distributed data of proportions should be centered around 0.5) in the meta-analytic model. So, we needed to transform the proportions through a double arcsine transformation method to address variance instability.
How to comment on this pooled estimate in the final paper, since even after the back transformation, the pooled estimate tends to be very different from the observed effect sizes of the included studies?
PS: We did the math through R." "
Thanks for your help!
Hello Mohamed,
The glib answer is, If the purpose of the meta-analysis is to report on the obtained ES values, then why worry about symmetry of the obtained proportions?
As well, there is good reason to be wary of the double arcsine transform applied to proportions, as outlined in this contribution: Article Seriously misleading results using inverse of Freeman‐Tukey ...
That is, other methods might be better applied and be less likely to yield "paradoxical" results via back transform.
Good luck with your work!
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Meta-Analysis