Logistic regression is in fact obtained by discretizing continuous variables with logistically distributed errors. This is the "threshold model". If you assume

Logistic regression is in fact obtained by discretizing continuous variables with logistically distributed errors. This is the "threshold model". If you assume normal instead of logistic you get the probit.