Using Restricted Maximum Likelihood in random-effects meta-analysis for binary outcomes

[Zad Chow]

I'm coming from R, so I'm hoping to find an easy way to get a similar output with Stata. I know that the default method is the Dersimonian Laird method but I'm hoping to be able to calculate a summary effect and confidence intervals using restricted maximum likelihood for odds ratios and maybe even risk ratios. I can't seem to find any commands that allow me to do so (like in the metaan command). I've only found some stuff indicating that there needs to be a variable for effect size and standard error/

[Alexander Rodriguez]

Hi Zad,
I have used metan a bit in the past. In the [_options] I'm sure you can specify rr for risk ratios. However, you may need raw count data to be able to do this.

For example you code may look like this:

metan events_exposure n_exposure events_control n_control, rr xlab(0.1, 0.5, 1, 5, 10) label(namevar=study, yearvar=year) fixed second(random) favours("Favours treatment" # "Favours control")

Hope this helps or a least gets you on your way.

Cheers, Alexander

[Zad Chow]

Thanks Alexander Rodriguez, but I don't think I see any place in the code where it specifies that the specific random-effects model being used is profile likelihood or REML

[Weiwen Ng]

Actually, you may want the -mvmeta- package instead, which does support REML (and ML, and method of moments)

https://www.stata-journal.com/sjpdf....iclenum=st0156

[John Moran]

See "admetan" by David Fisher: ssc install admetan