Efficient sequential designs for destructive life testing with application to animal serial sacrifice experiments

Abstract

We consider the scheduling of inspection times in a life testing experiment when testing is destructive. One application is to animal carcinogenicity tests where it is desired to find serial sacrifice schedules which allow efficient estimation of the onset time distribution for a tumour which is assumed irreversible, nonlethal and occult, i.e. detectable only at death. A class of designs are introduced and their properties are discussed. It is shown how to construct a design in this class that is asymptotically efficient. Tables of relative efficiencies are presented for the case of exponentially distributed failure times and test population sizes of 50 and 200. Some extensions of the theory are discussed.