Using the estimated covariance matrix, we proceeded with the GLS F-test to test the effect of each marker on the phenotype and then applied genomic control to quantify the amount of residual inflation. The genomic control λ parameters we obtained with EMMAX are much lower than those obtained using either standard association methods or regression analysis including 100 principal components (Table 1). Figure 3 and Supplementary Figure 2 illustrate the results using quantile-quantile plots of the P value distributions from these three tests. Only one of the ten phenotypes showed λ values within the 95% confidence interval of 0.992–1.008 with uncorrected or principal component analysis, whereas all of them fell in the confidence interval with EMMAX.
