What do you do with a categorical predictor (e.g. Black-White-Hispanic) in a logistic model when none of the cases are White, but about half the population studied is? You have to drop this predictor since the logit model cannot handle this, but you think it is important to talk about.... I was told "this is not a statistical relationship" but obviously there is strong evidence of a relationship. Indeed, the relationship would have to be weaker in order to be studied. There must be a name for this situation, and references discussing it, but I'm not sure how to look it up.
This would be like saying you can't prove women are more likely to go into labor than are men unless a man goes into labor.
