Since some of the area means did not have a variance (when they came from only one observation), we examined the effect of missing variances by fitting another model restricted to data sets that had non-missing variances.
We calculated missing variances from reported P values.
Among the multiple possible explanations for the "missing" variance, joint effects of multiple genetic variants are a particularly appealing target for investigation: they are well documented in biology and can often be evaluated using existing data.
U is evaluated as the posterior expectation of the realized score, and the observed information V is calculated by subtracting the missing information (posterior variance of the realized score) from the complete information (posterior expectation of the realized information).
Thus, C miss 2 is expected to underestimate the fraction of missing genetic variance.
Results for C miss 2 followed the same trends as for C miss but the values of C miss 2 were lower probably due to its underestimation of the fraction of the missing genetic variance.
For studies with missing measures of variance for mean change, SDs were estimated by using the correlation coefficient (r) method, in which the average correlation coefficient (r = 0.965) between baseline and trial-end values from all other included studies was used.
Estimates of C miss 2 differed from estimates of C miss, probably because C miss2 is expected to underestimate the fraction of the missing genetic variance.
To be conservative, we assumed a correlation of 0.5 to impute missing change variance estimates in the primary analysis.
