Some of you may be interested to read this new post.
https://bayesian-ma.net/2025/11/06/u...meta-analysis/
This is on the website associated with the book "Bayesian Meta-analysis" by me and Gian Luca Di Tanna.
The bottom line is that we show how to run meta-analyses as a special form of multilevel model in bayesmh and with the llevaluator() and evaluator() options. In fact , Stata 19 introduces complete flexibility in what you can send to and retrieve from [ll]evaluator(). That means you can sample from any distribution, using any scalars, variables, or matrices from Stata in your calculation of likelihood and prior. StataCorp have been a bit too humble about this in the marketing material, because it is a big deal. We also include a table in this post of the things that can be sent and retrieved, potentially inside each iteration of the algorithm, which opens up many computational possibilities as well as modelling ones. Some of that to and fro is undocumented in [BAYES] and could be useful to you.
https://bayesian-ma.net/2025/11/06/u...meta-analysis/
This is on the website associated with the book "Bayesian Meta-analysis" by me and Gian Luca Di Tanna.
The bottom line is that we show how to run meta-analyses as a special form of multilevel model in bayesmh and with the llevaluator() and evaluator() options. In fact , Stata 19 introduces complete flexibility in what you can send to and retrieve from [ll]evaluator(). That means you can sample from any distribution, using any scalars, variables, or matrices from Stata in your calculation of likelihood and prior. StataCorp have been a bit too humble about this in the marketing material, because it is a big deal. We also include a table in this post of the things that can be sent and retrieved, potentially inside each iteration of the algorithm, which opens up many computational possibilities as well as modelling ones. Some of that to and fro is undocumented in [BAYES] and could be useful to you.
