I have been doing GWAS on a scaled phenotype (from 0 to 16, where 0 represents extremely good, and 16 represents extremely bad). Also I have two covariates. Since scaled phenotype is not normally distributed, I binarized it into 0 and 1, and tried logistic regression with or without covariates using PLINK (v 1.9). I am quite confused about the --genotypic and --interaction parameters from PLINK though I had already read the PLINK manual. Logistic regression would give totally different results (p values of ADD or GENO_2DF) concerning these two parameters, and I also got inflated qqplots which squeezed the diagonal line almost horizontal. This inflated qqplot mainly occurred when I included --genotypic parameter into logistic regression, and I want to know why did this happen.