How to Achieve Stable Reliability Estimates: Effect of Instability in Resistance Data

The developers of early specifications for load and resistance factor design (LRFD) based many of their initial judgments on analyses of specific data sets representing structural elements of various types of materials. Because the specification committees based their final recommendations not only on analytical results, but also on calibration judgments, the resulting documents provided an excellent incremental step in the evolution of structural design. More recently, as reliability analysis procedures have become widely available, an increasing number of technical papers are discussing their results in terms of the reliability index of a given design. While some of the authors of these papers take particular care to disclose all of their underlying assumptions (some even conduct sensitivity studies on specific parameters), many authors do not. Specifically in the field of engineered wood products, some authors characterize their resistance data completely in terms of a single limited test program. Others do no testing at all, and rely on data from the historical literature as the basis of their analyses. This paper discusses reasons why specific "answers" in terms of specific reliability indices are difficult to obtain. It focuses on the magnitude of error that would occur if one mistakenly assumes that resistance data are static. The analysis in this paper focuses on the most stable calculation method — the format conversion option in ASTM D5457. Future work will explore the additional instability introduced when more advanced reliability calculation techniques are used. The paper concludes by recommending several techniques for dealing with resistance data variations for various classes of structural wood materials.