zhanxw / rvtests

Rare variant test software for next generation sequencing data
129 stars 41 forks source link

bgen logistic regression gives wrong results #72

Open winkusch opened 5 years ago

winkusch commented 5 years ago

Hi !

when running rvtests --single wald with UK biobank bgen files as input and binary oucomes, I am getting results that differ completely from other GWAS tools that allow to perform logistic regression with wald test and the bgen files.

The code I am using is rvtest --inBgen test.bgen --inBgenSample test.sample --pheno test.ped --pheno-name y12 --covar test.ped --covar-name age,sex01 --single wald --out test.out

Please note that the pheno file contains many missing values for the outcome and/or covariates.

Is this a known issue? I hope this can be solved soon - otherwise many analysts may produce erroro prone results.

Thanks!

Best, Thomas

winkusch commented 5 years ago

Small addition:

When adding --peopleIncludeFile test.samples to the command, this always gives a segmentation fault.

Is there an issue with loading samples from bgen files?

Best, Thomas

zhanxw commented 5 years ago

Can you please provide more details on input files:

test.bgen, test.sample, test.ped

If you can provide these files, I will debug to find out the exact problem.

winkusch commented 5 years ago

I have just sent the data to you via e mail. I hope it can be solved. Thanks!

zhanxw commented 5 years ago

Under investigation now. I will update here later.

zhanxw commented 5 years ago

Fixed in https://github.com/zhanxw/rvtests/commit/ff46d0af7ad49c5a81a2fad5a6a36801ff1b0854 and in master branch.