forest plots of ORs from categorical data using metan

[K MacQuarrie]

Hello,

I am using metan (SSC user written program) in Stata 14.1 to display forest plots from logistic regressions for about 10 separate surveys. No problems when I was displaying odds ratios for dummy variables. Now, I'd like to display forest plots of the ORs of a categorical variable (ref=never plus 2 categories, "sometimes" and "often").

My approach so far has been to graph one set of ORs for the "sometimes" category with the survey names in the left column:

metan logor loglci loguci, wgt(popweight) eform lcols(Survey)xlabel(0.05,10)force nowt title("DWSIN--Any IPV (sometimes)")

and then graph the second set of ORs for the "often" category with survey names:

metan logor2 loglci2 loguci2, wgt(popweight) eform xlabel(0.05,10)force nowt title("DWSIN--Any IPV (often)")

and then edit and combine the graphs into a small multiple set. 2 panels placed horizontally.

This feels like a clunky approach and I'm in search of an easier way. I'd gladly accept either a forest plot that has two panels, one for each of the categories (as produced by my clunky efforts above) or a single plot that has two "rows" of ORs for each survey.

Thank you for any advice!
Kerry

[Lise Bjerregaard]

Hi Kerry,
Unfortunately, I cannot help, but I have the same issue, so if you have found a solution in the meantime, I will be happy to hear that.
I would like to do an interaction test (likelihood ratio) based on the results from metan, but this requires that the meta-analyis in done simultaneously on all categories of my exposure, and it does not seem that metan can handle more than one at a time.
Lise