Predictor Sort Sampling and Confidence Bounds on Quantiles I. Asymptotic Theory;Forest Service research paper

摘要

Designers working with lumber must try to ensure that the strengths of wood structured members exceed the loads to which the members will be subjected. One approach to this problem is to design so that expected loads do not exceed allowable strength properties associated with particular species and grades of lumber (ASTM D1990). An allowable property is commonly obtained experimentally by taking a sample from the lumber population in question, obtaining a lower one-sided confidence bound on the fifth percentile of the strength distribution of the population, and then dividing by a safety and duration-of-load factor.