Logistic regression is frequently used for classifying observations into two groups. Unfortunately there are often outlying observations in a data set, who might affect the estimated model and the associated classification error rate. In this paper, the effect of observations in the training...