Figure 6.

Performance with Multiple Phenotypes

To illustrate the effect of combining more than one phenotype, we applied the Fisher type method (the sum of the log p values) to cholesterol and BMI regression coefficients. This figure shows the ROC curves when each one of the phenotypes was used compared to the curve when both were combined. Clearly, the combined method outperforms both single-phenotype methods. The AUC for each phenotype was 83% and 87%, whereas the combined AUC is 95%.