Parameter and quantile estimation for the distributions of failure strength of structural lumber

Estimating various characteristics, such as selected quantiles or the parameters, of a probability distribution of structural lumber failure strength is critical to quality control during manufacturing operations and for reliability-based structural design. This paper presents and compares the performances (in terms of bias and root mean square error) of several alternative estimation methods when censored strength data are used in the statistical estimation process. This study concludes that the method of maximum likelihood is nearly universally superior to a regression method that appears to be in wide use in the lumber industry. Recommendations are made about the use of alternative strategies for implementing the maximum likelihood method. The results of this work will be valuable when selecting the best estimation method for a particular set of circumstances. The results also establish the foundation for a computer program that will use a rule-based expert system to facilitate the estimation process.