search

Ripkelab / ricopili

Main ricopili repo for public releases
42 stars 23 forks source link

Imputation changes missing phenotypes to controls #30

Open rkwalters opened 8 years ago

rkwalters commented 8 years ago

In haps2dos4 lines 237-248, the fam file is processed to force all phenotypes to 1 or 2. The reason for this is unclear, but regardless this altered phenotype is propagated through the remainder of imputation.

As a result any individuals with non-case/control phenotypes, importantly including missing phenotypes of 0/-9, are recoded to cases. Unless an alternative phenotype is specified, this changed phenotype will also be used for GWAS.

stephanripke commented 8 years ago

thanks for this catch. first the individuals are recoded as controls (as you say in the header but not in the text). still this needs to be fixed. there was a reason for that back then, I think plink did not carry individuals with missing phenotype in the next dosage file. this might have been fixed meanwhile. I will fix it but it needs some testing. meanwhile you can use an alternate phenofile in postimp analysis. best, s.