In your documentation on logistic regression https://hail.is/docs/0.2/methods/stats.html#hail.methods.logistic_regression_rows
you point to a very interesting article: http://www.ncbi.nlm.nih.gov/pmc/articles/PMC4049324/ (thanks for this by the way)
Given my cohorts, the best recommendation is a joint analysis using Firth test. The appropriate Hail function to join the datasets seems to be
hl.MatrixTable.union_cols , is that right?
What happens when variants have an identical location in both sets to join, but have a different rsid?