Focusing on SNPs measured on one chip and imputed in the other, we considered removing SNPs with imputation R2 < 0.5,0.75,0.9,0.95 and 0.99. After thresholding by each value of R2, we calculated λ and the percentage of SNPs with p < 5×10−8. We kept track of the number of SNPs still available for analysis at each threshold.
