The failure time distribution for various items often follows a shifted (two-parameter) exponential model and not the traditional (one-parameter) exponential model. The shifted exponential is very useful in practice, in particular in the engineering, biomedical sciences and industrial quality control when modeling time to event or survival data. The open problem of simultaneous testing for differences in origin and scale parameters of two shifted exponential distributions is addressed. Two exact tests are proposed using maximum likelihood estimators. They are based on the combination of two statistics following a maximum-type and a distance-type approach. The exact null distributions of the respective test statistics are derived analytically. Small sample type-one error rate and power of the tests are studied numerically. We showed that the test based on the maximum type combination (the Max test) should be preferred being generally more powerful than the test based on the distance type combination (the Distance test). An application to a biomedical experiment is discussed.
This work deals with a general problem of testing multiple hypotheses about the distribution of a discrete-time stochastic process. Both the Bayesian and the conditional settings are considered. The structure of optimal sequential tests is characterized.