The measure does have problems eg https://t.co/tDw1RYAZDe but with a Cochrane median of 21% it’s probably safe to say this one is fishy.
RT @otis_reid: Are there meta-analyses in economics that don't have an I-squared of like 90%? In medicine (e.g. Cochrane library) it seems…
RT @otis_reid: Are there meta-analyses in economics that don't have an I-squared of like 90%? In medicine (e.g. Cochrane library) it seems…
Are there meta-analyses in economics that don't have an I-squared of like 90%? In medicine (e.g. Cochrane library) it seems like the I^2 is closer to 20% (https://t.co/B6PvtuO7FN), but most econ/social science meta-analyses have really high numbers.
@BallouxFrancois @whippletom It's substantially higher than the median estimate for Cochrane meta-analyses, and also biased downwards in truly heterogeneous datasets with small samples! https://t.co/nBpvIrnHWN
@cjsewall9 @dsquintana Yes, and https://t.co/sIsc56OmED Many of these seem to include very few studies (3 in one case it seems)...
The #heterogeneity #statistic #I2 can be #biased in small #meta-analyses; http://t.co/eM2Hlqc2fS #bmcmedicalresearchmethodology
The #heterogeneity #statistic #I2 can be #biased in small #meta-analyses; http://t.co/eM2Hlqc2fS #bmcmedicalresearchmethodology
"The heterogeneity statistic I 2 can be biased in small meta-analyses" http://t.co/EMp6M8VrPh
Estimates of heterogeneity (I2) can be biased in small meta-analyses. http://t.co/ZmMn257aSY