Testing the Koziol-Green model against monotone conditional odds for censoring

A notion of conditional odds for censoring is introduced to study the pattern of censoring in randomly right censored data. This notion is used to formulate meaningful alternatives in order to test the Koziol-Green model of proportional hazards between survival and censoring times. A test criterion is proposed and the test statistic is shown to be asymptotically normally distributed. The proposed procedure is illustrated through two data sets well known in the survival analysis literature.