OHDSI Home | Forums | Wiki | Github

Heterogeneity test in EvidenceSynthesis

Developers
#1

Hello all,

Recently I use EvidenceSynthesis packages when it comes to do meta-analysis.

However, when I use plotMetaAnalysisForest() function, it shows like below (it is an example image)

unnamed-chunk-13-1
unnamed-chunk-13-11400×865 27.4 KB

Seem like it shows tau statistics ans each study’s CI.

Is there any way to do Cochran’s Q and I^2 value tests in this pacakges?

If not, which method of packages to use?

Thank you :slight_smile:

#2

If you’re using Bayesian random-effect meta-analysis (which I highly recommend) then Q and I^2 are ill-defined. Furthermore, especially if you’re using non-normal likelihood approximations (again, highly recommended), these can be misleading metrics. Tau would seem a far more objective metric.

Any reason you prefer I2?

1 Like
#3

I’m sorry for the late. Does it means that I don’t need Q and I^2 value when it comes to using Baysian random-effect meta analysis? What exactly tau represent? I apologize my humble knowledge :frowning: Thank you!

t