Second, our sample was underpowered to detect the small effect sizes genes that are likely to impact on risk for AD. Looking at our larger EA sample and assuming an r2 of 0.8 between a SNP and a risk variant (Clayton, Chapman, and Cooper 2004), our power to detect a variant at the commonly accepted genome wide significance level of p ≤5 × 10-8 that influenced 0.5, 1.0., 1.5, 2.0 and 2.5% of variance in our AD factor was, respectively, 0.9%, 13.4%, 44.7%, 75.4% and 92.1%. We would have ≥ 80% power to detect a variant responsible for 1, 1.5 and 2% of trait variance at p levels of < 0.001, < 1.0×10-5 and < 1.0×10-6, respectively.
