Fig. 3.

Power of tests of disease association and heterogeneity of genetic effects between two sub-phenotypes, where the causal variant (MAF 10%) has a fixed effect on sub-phenotype 2. Results are presented as a function of the heterozygote log-relative risk of sub-phenotype 1 at a 5% significance level, where the causal variant has a heterozygote log-relative risk of 0.1 for sub-phenotype 2. MULTINOMIAL: test of association of the causal variant with disease, allowing heterogeneity of allelic odds ratios between sub-phenotypes, within a multinomial regression framework (2,000 cases against 2,000 controls). LOGISTIC: test of association of the causal variant with disease, assuming the genetic effect to be the same for both sub-phenotypes, within a logistic regression framework (2,000 cases against 2,000 controls). SP1: test of association of the causal variant with disease sub-phenotype 1 within a logistic regression framework (1,000 cases against 2,000 controls). SP2: test of association of the causal variant with disease sub-phenotype 2 within a logistic regression framework (1,000 cases against 2,000 controls). HETEROGENEITY: test of heterogeneity of the effect of the causal variant between sub-phenotypes within a multinomial regression framework (2,000 cases against 2,000 controls). SP1vSP2: test of heterogeneity of the effect of the causal variant between sub-phenotypes within a logistic regression framework (1,000 cases of sub-phenotype 1 against 1,000 cases of sub-phenotype 2).