Phasing alleles with SHAPEIT4: cases and controls together or separately?

Question

[igor.alcaras]

I have a merged VCF with SNP genotypes from 16 patients (cases) and 259 controls, and I need to phase their alleles. I'm using SHAPEIT4. But I don't know which decision to take.

1. Do I have to phase them separately, I mean, one phasing for patients and another for controls? I tried so and, as my case groupo is very small, I guess, many SNPs disappear and I need a great number...

2. For the other side, when I phase the VCF with both groups merged, all the SNPs remain there. But is it correct to phase both groups in the same VCF? Doesn't it causes vias in any possible haplotype associated with the phenotype I'm searching?

I really didn't find anywhere explaining the difference, neither which way is the correct.

Thanks for anyone who could give me this answer.