Patterns of long-term performance - How well are they predicted from accelerated tests and should evaluations consider parameters other than averages?
R C De Groot, J W Evans
This paper is a discussion of whether different service-life distribution patterns of products treated with unlike preservatives can be predicted, modeled, characterized, or even anticipated from accelerated laboratory tests. Graphic displays of data from Forest Products Laboratory field plots with preservative-treated and fire-retardant-treated stakes demonstrate the importance of local environment as a factor that affects field performance and exhibits differences in dose-response patterns among treatments. These distribution patterns are discussed with reference to early failures, first quartile and median failure times, and distribution about medians. Questions are then asked about the relevance of these parameters to practical applications, about the need to consider population characteristics other than average in evaluations on new preservatives, and about the capability of accelerated tests to estimate these parameters.