QPRC 2016

Planning Accelerated Life Tests with Random Effects

Rong Pan and Kangwon Seo
Arizona State University

Randomization is one of basic principles of experimental design, which assumes that all experimental responses are independent to each other. Planning an accelerated life test (ALT) with multiple stress factors can be treated as an experimental design problem, thus it should follow this principle. In reality, however, the randomization principle has to be violated because of the engineering and economic constraints imposed on ALTs. Typical experimental protocols in ALTs include subsampling or random blocks, and both lead to a grouped structure among observations. In this talk, we will demonstrate how to use a generalized linear mixed model (GLMM) approach to analyzing multiple stress ALT data with constrained randomization. We will derive optimal ALT test plans for this kind of model and compare their statistical properties with those of conventional ALT test plans.