I'm trying to run a fixed effects meta-analysis using the Metafor R package and in the documentation I read that it does not return an I^2 for fixed effects models, only for random effects models. This is because the I^2 is calculated using tau^2, which is only computed for random effects models. While the tau^2 is 0 for all the fixed effects models, as it should be, it's giving me non-zero I^2 values.

Does anyone know why it's giving me an I^2 greater than 0 for fixed effects models?

Here's my R macro for the model:

metaFE <- defmacro(data, subset,
               expr=rma(as.numeric(Beta), sei=as.numeric(StdErr), measure="RR", slab=as.character(study), data=data, subset=subset, method="FE"))

and here is the formula Metafor uses for I^2:

I^2 = 100% × [tau^2 / (tau^2 + s^2)]

Any insight is appreciated! Thanks!

  • 1
    $\begingroup$ This might get a better answer on the mailing list stat.ethz.ch/mailman/listinfo/r-sig-meta-analysis/ You need to register there first. But you will need to give much more detail there as this is not reproducible as (a) we do not have your macro, (b) we do not have your data, (c) we do not see your output. $\endgroup$ – mdewey Jul 23 '19 at 15:45