Across all three SNP panels, adjusting for nongenetic effects using PC generated threefold more eQTLs than the unadjusted analysis. For either adjusted or unadjusted expression data, imputing 2.4M HapMap SNPs gained 6%–7% additional signals, while imputing the ∼8M 1000G SNPs further increased the number of signals by 5%–8% (a total increment of 12%–15%). Power gain from increased marker density thus appeared to be largely independent of the beneficial effects of reducing measurement error.
