Score computation for linear rank procedures

Abstract

Efficient scores for linear rank two sample procedures on censored data take the form of multiple integrals which often cannot be expressed in closed algebraic form. In this paper we show that efficient scores for censored data can be written as a linear combination of the efficient scores for uncensored data. A recursive scheme for computing the coefficients in these linear combinations is given. The usefulness of the procedure is demonstrated by its application to censored lognormal data. The accuracy and applicability of the algorithm is discussed.