search

AbbVie-ComputationalGenomics / SAIGEgds

Scalable Implementation of generalized mixed models using GDS files in Phenome-Wide Association Studies
7 stars 5 forks source link

Results: beta (odds ratios) < 0 #10

Closed SunYidan2021 closed 1 year ago

SunYidan2021 commented 1 year ago

In the result, "beta" indicates : beta coefficient, odds ratio if binary outcomes (alternative allele vs. reference allele). However, when I conduct GWAS on binary outcomes, there are some SNPs with beta < 0. It's impossible that the odds ratio is < 0. Would you mind me asking how to interpret it? Should I delete SNPs with beta < 0? 

   id chr      pos rs.id ref alt       AF.alt   mac   num        beta         SE       pval
1:  3  22 16868201         C   A 0.0004218781    17 20148  0.95400134 1.41909658 0.50141843
2:  5  22 16868590         G   A 0.0005707763    23 20148 -1.06041310 1.18949813 0.37267203
3:  6  22 16868603         C   T 0.0003226127    13 20148 -1.06501842 1.50172130 0.47820133
4:  7  22 16868711         G   A 0.0002977963    12 20148 -1.05899912 1.67452636 0.52711436
5: 12  22 16869223         G   A 0.2621351995 10563 20148 -0.12863379 0.06312118 0.04156131
zhengxwen commented 1 year ago

beta is log odd ratio (can be < 0)!