A shrinkage type estimator is introduced which has favorable properties in binary regression. Although binary observations are never very far away from the underlying probability, in all interesting cases there is a non-zero distance between observation and underlying mean. The proposed response smoothing estimate is based on a smoothed version of the observed responses which is obtained by shifting the observation slightly towards the mean of the observations and therefore closer to the underlying probability. Estimates of this type are very easily computed by using common program packages and exist also when the number of predictors is very large. Moreover, they are robust against outliers. A combination of response smoothing estimators and Pregibon's resistant fitting procedure corrects for the overprediciton of the resistant fitting in a very simple way. Estimators are compared in simulation studies and applications.
Logit model resistant fitting response smoothing estimator shrinkage weighted estimation data sharpening