Testing for the Proportionality of Hazards in Two Samples Against the Increasing Cumulative Hazard Ratio Alternative (Q4702182)

The proportional hazards (PH) model has played an instrumental part of data analysis in such areas as survival analysis, reliability, economics, demography and environmental studies. The validity of the PH assumption in a two-sample problem may be checked through one of the traditional graphical methods proposed by \textit{D.R. Cox} [J. R. Stat. Soc., Ser. B 34, 187-220 (1972; Zbl 0243.62041)], \textit{P.K. Andersen} [Biometrics 38, 67-77 (1982; Zbl 0479.62081)], etc. In recent years researchers have proposed tests geared especially for the alternative hypothesis of ``increasing hazard ratio'', keeping in mind the case of crossing hazards. In this paper a test of the proportional hazards model for the weaker ``increasing cumulative hazard ratio'' alternative is developed. The work is motivated by a data analytic example given by \textit{R. Gill} and \textit{M. Schumacher} [Biometrika 74, 289-300 (1987; Zbl 0621.62047)] where their test fails to reject the null hypothesis even though the faster ageing of one group is quite apparent from a plot. The normalized test statistic proposed in this paper has an asymptotically normal distribution under either hypothesis. Two graphical methods related to the analytical test are also presented.