Skip to content
Prev 3268 / 5632 Next
Original Raw

[R-meta] I2 or p to choose the effect model

Wolfgang Viechtbauer 
Hi Lorenzo,

I^2 is a relative measure of heterogeneity. Saying 40% (or whatever cutoff) indicates 'great heterogeneity' is therefore not correct. 40% could represent a trivial amount of heterogeneity (in an absolute sense) if the studies are all quite large or it could suggest huge fluctuations in the underlying true effects if the studies are all quite small (or anything in between).

Borenstein discusses this in his books. See also:

R?cker, G., Schwarzer, G., Carpenter, J. R., & Schumacher, M. (2008). Undue reliance on I2 in assessing heterogeneity may mislead. BMC Medical Research Methodology, 8(79). https://doi.org/10.1186/1471-2288-8-79

Also, choosing a model based on the presence/absence of a certain amount of heterogeneity (however quantified) is not really a defensible strategy. It really comes down to the types of inferences you want to make. See here:

https://wviechtb.github.io/metafor/reference/metafor-package.html

Best,
Wolfgang

Thread (2 messages)

Martin Lobo I2 or p to choose the effect model Sep 15 Wolfgang Viechtbauer I2 or p to choose the effect model Sep 15