Discussion of: Asymptotically Optimum over Stress Tests to Estimate the Survival Probability at a Condition with a Low Expected Failure Probability

Abstract

Drs. Meeker and Hahn, in examining optimal overstress tests for the logistic quantal response model, have provided useful experimental design plans, as well as some insights into the general nature of the optimal designs, and the effect of the design plan on the variance of estimates. My discussion will deal mainly with practical questions relating to the use of these designs. Of particular interest are questions concerning the effect of departures from the assumed model upon the test plans, and more generally upon the inferences which can be drawn from data of the type envisaged here. Drs. Meeker and Hahn carefully point out the exact assumptions upon which their optimal plans rest, but do not examine in any detail the effect of departures from these assumptions, so that these comments should serve as a practical supplement to their work. The mathematical details concerning optimal design of experiments for different quantal response models are similar, and I will begin in the next section by outlining these. This will be of help later when we wish to examine the effects of departures from the logistic model assumed in the paper.