Open prayaggordy opened 5 months ago
For cat-cat interactions, ABCs require joint probabilities. E.g., for sex:race, it's not enough to look at the sample proportions for Male and Female, and then for NHW, NHB, H, etc.; we have to use joint proportions. For ABCs that use the observed sample proportions, colMeans(X) actually gives that to us quite easily. But this does not generalize for when a user wants to input their own proportions (e.g., known population proportions, or all 1's to do sum-to-zero constraints).
Two "FIXME" notes in
getConstraints()
. Dr. Kowal, could you remind me what the issue was?