The research interests of the investigator deal with various aspects of sequential analysis. Sequential survival analysis makes it possible to construct sequential tests of known size and power under the frequently used Cox proportional hazards model for the data. As a special case of sequential experimental design, one can find the drug dose which will produce a certain response with a specified probability and find the stress level which will cause a specified percentage of components to fail. Observed significance levels for tests of hypothesis are commonly used as judging criteria by the frequentists. Bayesian analysis for lower bounds on the evidence against the null hypothesis shows that the observed significance levels typically give a very misleading impression as to the actual evidence in the data. Professor Sellke has already published several results in all of the above areas. He will continue his work in them.
