Key points are not available for this paper at this time.
Probit analysis can be0 reformulated in a model where the subject's response or non-response is determined by which one of two random variables, indicants, is the greater. The extenstion of this formulation to more than two categories of response leads to a new generalization of probit analysis, which raises interesting identinability and estimation problems.
Aitchison et al. (1970) studied this question.