Semiparametric analysis of discrete response. Asymptotic properties of the maximum score estimator (Q1059977)

This article considers the estimation and identification of the model \(MED(y| x)=x\beta\) from a random sample of observations on (sgn y, x). The author has shown that the parameter \(\beta\) is not identifiable (up to a scale factor) if the distribution of x has bounded support and the median regression function \(x\beta\) is uniformly bounded away from zero for all x. Identification can be achieved under infinite support conditions. Least absolute deviations estimators are shown to be strongly consistent. Maximum score estimation is shown to be equivalent to the least absolute deviations estimation procedures. The author has also shown that the maximum score estimate lies outside any fixed neighborhood of the normalized true parameter vector with probability that goes to zero at exponential rate.