CENSORED REGRESSION - LOCAL LINEAR-APPROXIMATIONS AND THEIR APPLICATIONS

Various statistical tools are available for modeling the relationship between response and covariate if the data are fully observable. In th e situation of censored data, however, those tools are no longer direc tly applicable. This article provides an easily implemented methodolog y for modeling the association, based on censored data. The form of th e regression relationship will be completely determined by the data; n o assumptions are made about this form. Basic ideas behind the methodo logy are to transform the observed data in an appropriate simple way a nd then to apply a locally weighted least squares regression. The prop osed estimator involves a variable bandwidth that automatically adapts to the design of the data points. That the methodology is very easy t o implement is illustrated by several examples, including simulation s tudies and an analysis of the Stanford Heart Transplant Data and the P rimary Biliary Cirrhosis Data. Several theoretical considerations are reflected in the examples. Finally, some basic asymptotic results are established.