The standard error of the estimate is se(bXY)=(βkXTΩ−1βkX)−1 [27]. After pruning for SNPs with linkage disequilibrium r2 > 0.8, all variants with a PeQTL < 5 × 10−8 in the cis-eQTL region of each probe available in both BMD and eQTL data were included in the analysis. The correlation between IVs was calculated in Plink software (version 1.0.7) by using HapMap 2 CEU as a reference panel. Cochran’s Q statistic was performed to test heterogeneity using the meta R packages [27], a p value > 0.05 can be interpreted as no more heterogeneity between causal effects estimated using the variants individually than would be expected by chance.
