Hi,

I am doing meta-analysis for individual patient data. I get different overall results when I use [1] ipdmetan and forestplot (ssc) or [2] meta set and meta forestplot (Stata v16.1). I can see that the weights values are different between the 2 methods; but I don't know why, as both should be using inverse-variance weighting. Should I add any option so that both methods yield to similar pooled estimates?
Please see below my syntax.
Code:
ipdmetan, study(study) eform saving(Data\temp\bin`i'.dta, replace): glm bin01 i.pred1 i.pred2, fam(bin) link(log) eform
Then I use the saved estimates to get the forest plot.
Code:
forestplot estimate min95 max95,                 ///
                                dp(2) effect(Crude RR)                                     ///                    
                                nostat lcols(_meta_studylabel IST IPT) rcols(RR95CIp wgt) ///
                                favours("{bf:s better}" "<{hline 10}" # "{bf:p better}" "{hline 10}>")     ///
                                graphr(col(white)) null(1) cirange (0.5 2.5) range(0.5 2.5) astext(85) aspect(0.15)                     ///
                                leftjustify xlab(0.5 1 1.5 2 2.5) nowt nobox graphregion(color(white)) bgcolor(white) noadjust
Code:
meta set estimate min95 max95, civartolerance(.5) fixed
 meta forestplot, subgroup(outcomes)
Thanks for your help
Carole