Abstract
When genetic markers are in complete linkage disequilibrium with disease loci, it has been shown that the efficiency robust tests, including maximum (minimum) type statistics and the procedure with genetic model selection, are often preferred to either a single trend test or Pearson's chi-squared test in candidate-gene studies and genome-wide scans. We consider a more general two-locus model where linkage disequilibrium is allowed to vary, and examine the impact of penetrances at the marker locus when genetic models are defined at the disease locus. Robust statistics are then reviewed and their efficiency and robustness are compared through simulations in genome-wide scans of 300,000 markers under the two-locus model. Applications of the robust tests to Wellcome Trust Case-Control Consortium (2007) are illustrated.