Imputation changes missing phenotypes to controls

[rkwalters]

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]

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.