Statistical analysis of durability tests, part 1: principles of distribution fitting and application on laboratory tests

Service life prediction is an important topic in wood research, especially with regard to the Construction Products Regulation (CPR). Both laboratory tests as well as in-service performance testing are therefore essential in combination with proper monitoring and analysis tools. A crucial concept is variability in testing and analysis, especially for a biological material such as wood. The larger the sample size the more representative this is for the entire population, yet the number of specimens is often limited by a financial upper limit. Therefore it is essential to use the sub-optimal amount of data and assess as accurately as possible the characteristic under study. In this paper we focus on the use of probability density functions (pdf), also known as distributions. The principles and guidelines for pdf fitting will be explored as well as the use of confidence intervals. The theoretical concepts will be applied on mass loss data. Intra- and interspecies variability but also inter-laboratory variability is illustrated. Therefore the analysis of test results of a round-robin as described in Brischke and co-workers (2013) will be illustrated as well as the analysis of lab tests performed at Woodlab-UGent according to CEN/TS 15083-1 (2005). A validation procedure, as part of a future updated standard, can be useful to erase inter-laboratory differences. Furthermore, the use of a reference wood species can also be an option as a benchmark to compare other species rather than using ‘absolute’ testing resulting in a ranking based on median values. In Part 2 of this paper we will then further use the concepts of pdf fitting for time-to-failure analysis of field test data