A Diagnostic Statistic for the Score Test
本文推导了删除观测值时得分检验变化的精确和近似表达式，并应用于斯坦福心脏移植数据，帮助识别对检验结果影响大的个体或层。
Abstract This article provides exact and approximate expressions for the change in the score test on deletion of observations. Deletions of individuals and of entire risk sets from matched case-control and survival studies are discussed in detail. A simple formula is derived, which measures the contribution of individual strata to score tests based on combining several contingency tables. The diagnostic is illustrated by application to the Stanford heart-transplant data.