The conventional alpha or Type I error level to declare a finding significant (p<0.05) was corrected for multiple tests. Because the analyzed SNPs are in LD with each other, the tests are not independent. Thus, the method proposed by Nyholt (Nyholt, 2004) was used to evaluate the correlation between SNPs to determine the effective number of independent tests in the analysis. To do this, we used Haploview to calculate the number of haplotype blocks included in the candidate gene using the procedures described below. Tagging SNPs that did not fall within a defined block were counted as an independent test. We estimated a total of 8 effective independent tests. Therefore, using a conservative approach, we applied a Bonferroni correction for Type I error for 8 tests (α = 0.05/8=0.0062). The tests were two sided with a 0.0062 significance level.
