Shrinking individual study estimates in meta-analysis

[paulvonhippel]

Meta-analysis summarizes the results of individual studies. There are empirical Bayes methods for shrinking the estimated effect of each study toward the mean effect across studies. Are these shrunken estimates available in Stata today? From this post I understand they were available in an old user-written command called meta, but that was supplanted when Stata 16 published a native command with the same name.

What are the options today?

Thanks!
Paul

[Tiago Pereira]

Paul, I am not aware of any package in Stata that can compute those shrunken estimates.

I am not entirely sure the meaning of "shrunken estimates", but I think you mean the posterior distributions for the study-specific effects.

It is possible to fit an OpenBUGS meta-analysis model from Stata and get those posterior distributions.

[David Fisher]

paulvonhippel This functionality is available within user-written metan (from SSC; latest version 4.08.1 12jul2024). The option is named ebshrink.

[Tiago Pereira]

As usual, thank you, David!

Is there an official publication mentioning that functionality?

[David Fisher]

Thankyou! Not as yet, I'm afraid... I have been meaning to submit an update to Stata Journal for ages, but never manage to find the time.

[paulvonhippel]

Thank you, David Fisher! I do see the option ebshrinkage documented in help metan.

Best,
Paul