Using our type 2 diabetes example in CaTS, Figure 1b shows the sample sizes required, in a one-stage design, to detect different combinations of disease allele frequency and GRRs with 80% power, assuming a SNP-based significance threshold of 1 × 10−7 (using a Bonferroni correction for all 500,000 tagSNPs on the panel) and of 1.67 × 10−7 (when assuming that 500,000 tagSNPs on the panel correspond to 300,000 effective independent SNPs). A control:case ratio of 1:1 and a multiplicative disease model (GRRAA=(GRR)Aa2) were assumed.
