".glm.firth.hybrid" output file in plink2 does not give correct CASE_ALLELE_CT values

[metzge...@gmx.de]

Hi, I have been doing a GWAS with PLINK2 and got an output file with the extension: .glm.logistic.hybrid command. Everything worked fine until I realized that the output values are not really conform with my phenotype values.

For example: For each sample I got phenotypes in binary format (0's and 1's). Here I have 347 samples with 0's (controls) and 194 with 1's (cases).

The output from PLINK2 gives me within the column CTRL_ALLELE_CT (specified by plink2 index as: Control allele observation count): 598. 

And in the column: CASE_ALLELE_CT (specified by plink2 index as: Case allele observation count): 304

Obviously these values in the ".glm.firth.hybrid"-file are not the correct case and control counts I inputted from the phenotype file. How can this be? What exactly is PLINK2 measuring here?

[Christopher Chang]

Three possibilities:

1. Missing covariate values.  Any sample with a missing covariate value is excluded from the analysis.

2. Single-sample categories.  If there's a covariate that's zero-valued for all but 1 sample, there is no point including that 1 sample or the covariate in the regression, and plink2 automatically excludes the sample-covariate pair.

3. Missing genotypes.

You can use --glm's "pheno-ids" modifier to see what samples remain after the first two considerations.