A generalized regression model for a binary response

In most cases authors are permitted to post their version of the article (e.g. in Word or Tex form) to their personal website or institutional repository. Authors requiring further information regarding Elsevier's archiving and manuscript policies are encouraged to visit: a b s t r a c t Logistic regression is the closest model, given its sufficient statistics, to the model of constant success probability in terms of Kullback–Leibler information. A generalized binary model has this property for the more general φ-divergence. These results generalize to multinomial and other discrete data.