On testing for a constant hazard against a change-point alternative.

A frequently recurring question posed by leukemia researchers concerns a test of a constant failure rate against the alternative of a failure rate involving a single change-point. In answer to this question, a likelihood ratio test appropriate for the stated alternative is derived and simulated. Consideration is given also to tests based on alternatives in the log gamma family, which perform quite well when the change-point model is correct. A practical application is given.