从子组的输出中提取元分析估计和置信区间

Question

I am trying to do almost exactly this thing: Extracting meta analysis estimate and CI from output in R

I have metaprop output, and I'd like to extract the confidence interval estimates to put into a dataframe so I can add them to my graph here: Forest Plot, ordering and summarizing multiple variables As a line for summary data.

I've used the double arcsine transformation and so the code listed in that first answer works perfectly if I'd like to get the esimate and CI for the plot as a whole, but I have subgroups. So how would I extract the numbers in green in the photo below, not the red one?

Edit: Update with more information, so per the SE question cited above, you do:

random.est2 <- c(pes.summary$TE.random,pes.summary$lower.random,pes.summary$upper.random)
unlist(lapply(random.est2, meta:::backtransf,  sm="PFT",  n=1/mean(1/pes.summary$n)))

To get the information out for the entire thing. Ie that code above gives me 0.988492 [0.976040, 0.997163]. These numbers line up perfectly with whats in the red square in the photo. And so its the "TE.Random" object that you extract for that. Well... per the documention for metaprop, "Te.Random.w" should be the estimated treatment effect for the subgroups. So I figured I'd do the same code as above for the subgroups and tried:

random.est2 <- c(pes.summary$TE.random.w,pes.summary$lower.random.w,pes.summary$upper.random.w)
unlist(lapply(random.est2, meta:::backtransf,  sm="PFT",  n=1/mean(1/pes.summary$n)))

But this gives me numbers that are roughly .01 off of what they should be. So I'm assuming maybe I'm backtransforming them incorrectly?

Lastly, there seems to be an n.harmonic.mean.w which is: "Harmonic mean of number of observations in subgroups (for back transformation of Freeman-Tukey Double arcsine transformation) - if byvar is not missing."

That sounds even more like what I should use, but I can't get that to work either.

Answer 1

Try this.

library(metafor)

Estimate<-transf.ipft(pes.summary$TE.random.w,ni=pes.summary$n.harmonic.mean.w)
Upper<-transf.ipft(pes.summary$lower.random.w,ni=pes.summary$n.harmonic.mean.w)
Lower<-transf.ipft(pes.summary$upper.random.w,ni=pes.summary$n.harmonic.mean.w)


new.df<-data.frame(Estimate,Upper,Lower)

For more information, see https://www.researchgate.net/publication/325486099_How_to_Conduct_a_Meta-Analysis_of_Proportions_in_R_A_Comprehensive_Tutorial