The assessment of goodness-of-fit for logistic regression models using categorical predictors is made complicated by the fact that there are different ways of defining the saturated model. Three ...