Nested group meta-analysis including random effects

[LATTERINIFRANCESCO]

Dear Developers of MetaWin,
first of all I would like to congratulate with you for the very great software and to thank you very much for providing such a tool.
However, I am asking if, and if yes when, do you think it will be possible to have the possibility of adding random effects to nested group meta-analysis. I am a researcher in forestry sector and when conducting meta-analysis it is rather difficult to have input data for meta-analysis which are not somehow nested (for different studies sharing the same control or the same group of authors conducting several researches on one topic). Therefore to have the possibility of nested group analysis but including random effects is very important, considering that applying a fixed-effects model can be reductive.
Thank you in advance and kindest regards
Francesco Latterini

[msrosenberg]

I'd love to do this, but the problem is there is no way to estimate pooled variance for a nested meta-analysis with random/mixed effects using a least-squares/moments estimator.

It can be done using maximum likelihood and Bayesian approaches, but I've run into technical problems implementing each approach so far. This is largely an issue of what is available within Python while also being truly portable and distributable across operating systems (e.g., many of the Bayesian solvers are not truly multi-platform and do not work within Windows without tremendous difficulty, making them unacceptable general solutions for the problem).

So this should be possible, but the timeline is unknown until we can solve some of these technical problems.

[LATTERINIFRANCESCO]

Thank you very much for the answer and congratulations again for the software and your efforts. Kindest regards Francesco Latterini Il mer 21 set 2022, 15:49 Michael Rosenberg ***@***.***> ha scritto:

…

I'd love to do this, but the problem is there is no way to estimate pooled variance for a nested meta-analysis with random/mixed effects using a least-squares/moments estimator. It can be done using maximum likelihood and Bayesian approaches, but I've run into technical problems implementing each approach so far. This is largely an issue of what is available within Python while also being truly portable and distributable across operating systems (e.g., many of the Bayesian solvers are not truly multi-platform and do not work within Windows without tremendous difficulty, making them unacceptable general solutions for the problem). So this should be possible, but the timeline is unknown until we can solve some of these technical problems. — Reply to this email directly, view it on GitHub <#6 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/A3GEJ2JHZUDX6Q74K7PBRF3V7MG73ANCNFSM6AAAAAAQRWXKXY> . You are receiving this because you authored the thread.Message ID: ***@***.***>