I made a metaregression with "metafor" package for R, and I would like to make a bubble plot too to show the direct relationship between the predictor and the outcome. Unfortunately, I still haven't found that function in the package "metafor": does anyone know ho to solve this problem? Thanks a lot! :D
You can actually find a nice sample of code on the official metafor JSS article, page 17:
http://core.ac.uk/download/pdf/6340274.pdf
Essentially, just plot effect size as a function of your moderator variable, but use the cex command to adjust the size of the points accordingly by the sampling variances. The plotrix package function "rescale" is a nice way to adjust the circle size, in that you could take your sampling variances and rescale them to values between 0.5 and 2.
So in practice, this would look like:
library(plotrix)
size=rescale(1/data$vi,c(0.5,2))
plot(yi~moderator,data=data,cex=size,las=1)
Where yi and vi are the standardized effect sizes and sampling variances calculated with escalc. And you can modify the labels and such accordingly. As noted in the JSS article, you can also add the model fit and confidence intervals through the predict.rma function.
I hope this helps!
Thank you Daniel, you gave me very interesting suggestions I will immediately try!
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Statistical Software
- R