In GWAS, it is standard to use (unphenotyped) cohort or population samples in place of control samples but analyze it as a case-control study using logistic regression. This is actually equivalent to fitting a log risk regression model [Schouten et al., 1993]. Thus, log risk regression is an appropriate model to consider in this context. The two models are related analogously to the way that the odds ratio (OR) and relative risk (RR) are related, and will be approximately equivalent when the disease prevalence is relatively small.
