Least Squares Method in Heteroscedastic Censored Regression Models.

摘要

Consider the heteroscedastic polynomial regression model Y=Beta(sub 0) + Beta(sub 1)X++Beta(sub p)X(sup p)+square root of Var(Y(straight line)X)epsilon where epsilon is independent of X, and Y is subject to random censoring. Provided that the censoring on Y is 'light' in some region of X, the authors construct a least squares estimator for the regression parameters whose asymptotic bias is shown to be as small as desired. The least squares estimator is defined as a functional of the Van Keilegom and Akritas (1999) estimator of the bivariate distribution P(X less than or equal to x, Y less than or equal to y), and its asymptotic normality is obtained.