Using the method proposed by Yang et al. (2010), the amount of variance in AD risk that is explained simultaneously by genome-wide SNPs was estimated by treating the effects of SNPs as statistically random. The model for this analysis is y = Σ wibi + e, where y is the phenotypic value, bi is the effect of the ith SNP, wi is a scaling factor equivalent to (xi − 2pi)/(2pi (1 − pi))1/2 with pi the allele frequency and xi the genotype indicator of the ith SNP (values of 0, 1 or 2), and e is a random environmental effect (Visscher et al. 2010). In matrix notation this is equivalent to y = g + e, where g = Wb is a vector of genetic values calculated from the SNP alleles each individual carries, with var(g) = WW′σb2 (WW′ is the matrix of genetic relationships between individuals). Using the software GCTA (Yang et al. 2011), we computed the genetic relationship matrix (GRM) for our LD-pruned genotype data, combining the COGA and SAGE samples for the EA (n = 2,763) and
