| Correct identification of the important genetic markers in a genome-wide association study is a formidable task due to the high chance of false positives. This work focuses on the problem of identifying the subgroups and their associated markers that yield large absolute risk reductions in a placebo-controlled setting. Several promising methods are proposed and examined for detecting subgroups defined by two or more of such markers. The methods employ decision trees, importance scoring, LOWESS smoothing, multi-step searching, screening and random data perturbation. Results from simulation experiments demonstrating the effectiveness of the methods are reported. |
