Results were adjusted for multiple testing by using the R package QVALUE v1.1 (Storey, 2003, Storey & Tibshirani, 2003). Rather than dealing with the probability of rejecting a true null hypothesis (the family-wise error rate), the approach implemented in QVALUE estimates the expected proportion of false positives among all rejected hypotheses (the false discovery rate, FDR) (Benjamini et al., 2001). QVALUE takes a given set of P values and, for each test, estimates the minimum FDR that is incurred when calling that particular test significant (the q-value of the test). The q-value measures the significance of each of a family of tests performed simultaneously and holds under different forms of dependence. The smallest nominal P value of all tests performed (Pmin) was observed from the haplotype association tests. The QVALUE program was run on the list of P values created by adding Pmin to the set of P values obtained from the single-locus tests. The result is the estimated experiment-wise significance of Pmin. An FDR of 0.05 was used as the significance level.
