Bayesian meta-analysis - which coefficients for effect size can be used?
Hello all,
I've been using JASP for sometime and decided to do a Bayesian meta-analysis in there. However, I am wondering which effect size can be used. I have Pearson's correlations with 95%CI (lower/upper bound) - does it make sense using them as input?
Best,
Vukasin
Comments
Hi Vukasin,
Great question. Anything will work, just as long as the prior distribution still makes sense. For instance, on our JASP post about this (https://jasp-stats.org/2020/05/12/how-to-conduct-a-bayesian-model-averaged-meta-analysis-in-jasp/) we initially used Relative Risk; however, our prior distribution is symmetric around 0, the value implicitly specified by our null hypothesis. So we used log(RR) instead.
Cheers,
E.J.