Multiple linear regression analyses of LIPG T111I SNP in the bi-ethnic population samples of San Luis Valley in Colorado.

Multiple linear regression analyses of LIPG T111I SNP. The analyses were performed separately for Whites and Hispanics. The model is adjusted for gender, age, BMI, and smoking. The regression models were performed first for significant covariate(s) selection using stepwise variable selection. Then T111I SNP was added to the model that included significant covariates in an ‘enter’ method. Both genotypic and allelic association of the T111I with the lipid panel measurements were tested. For genotypic association, the genotypes were recoded in order to give weight to each genotype as following: CC = 0, CT = 1, TT = 2. For allelic association studies, dummy variables were created for the C/T alleles as following and both were included in the regression model: D_TT = 0 (for T/C and C/C genotypes); D_TT = 1 (for T/T genotype); D_TC = 0 (for T/T and C/C genotypes); D_TC = 1 (for T/C genotype). Similarly, dummy variables were created for covariate ‘smoking’ as following: D_Smoker = 0 (for nonsmokers and ex-smokers) and D_Smoker = 1 (for current smokers). B-value is the regression coefficient, which indicates the amount by which the tested variable is changed. The details of each significant linear regression model are provided under asterisk (*).