Post by Jian Yang on Sept 24, 2015 22:15:52 GMT
If there are multiple cohorts and for some reason you are unable to pool all the individual-level genotype data together for a combined analysis, then it is OK to run a inverse-variance meta-analysis, i.e.
h2meta = sum(h2i / SE2i) / sum(1 / SE2i)
However, this is less powerful than a combined analysis because the meta-analysis does not utilise the contrasts between individuals across cohorts.
h2meta = sum(h2i / SE2i) / sum(1 / SE2i)
However, this is less powerful than a combined analysis because the meta-analysis does not utilise the contrasts between individuals across cohorts.