I ran a GWAS analysis based on logistic regression model that involves the interaction between SNP additive effect and an exposure using probABEL. To check the results I ran the same thing in PLINK. Some of the results differ (e.g. coefficients of some SNPs differ between probABEL and PLINK, and can be quite much; while those just differ by <0.01 were considered the same). I wonder if this is normal to happen? Could someone explain why some of the results differ? I just hope I ran the analysis correctly.