For the disease traits, the odds ratios of the haplotypes were evaluated by the haplotype-specific logistic regression model [27], and the amount of liability variance explained by the haplotypes was calculated using the standard liability threshold model [29]. The statistical significance of the variance was assessed using a likelihood ratio (LR) test between the null model and the full model with the haplotypes; the covariates of gender, age, age-squared, smoking cessation status (former/current) and the first and second PCs for the East Asian population were used in both of the models.
