作者: Andreas Christmann , Peter J Rousseeuw
DOI: 10.1016/S0167-9473(00)00063-3
关键词:
摘要: In this paper, we show that the recent notion of regression depth can be used as a data-analytic tool to measure amount separation between successes and failures in binary response framework. Extending algorithm, allows us compute overlap data sets which are commonly fitted by logistic or probit models. The is number observations would need removed obtain complete quasi-complete separation, i.e. situation where parameters no longer identifiable maximum likelihood estimate does not exist. It turns out often quite small. results equally useful linear discriminant analysis.