The focus of this talk is on failure history of repairable systems for which the relevant data comprise successive event times for a recurrent phenomenon along with an event-count indicator. Such data commonly occur both in industrial and biomedical contexts. In an industrial setting, typically a single expensive prototype of a complex system is observed until failure, followed by corrections or design changes, and subsequent retesting. In the biomedical context, however, it is more common to encounter data on multiple subjects with only a few recurrences per subject. In this talk we will describe the findings from a study of failure data both from a single repairable system and multiple repairable systems to multiple failure modes, a framework traditionally dubbed as competing risks. We adopt a parametric premise and discuss the results under the Power Law Process model that has found considerable attention in describing recurrent hardware failures of complex mechanical systems. Some interesting and non-standard asymptotic results ensue in this context that will be discussed in detail. We will report findings from an extensive simulation study that supplements the theoretical findings. The methodology will be illustrated on recurrent failure data obtained from a warranty claim database for a fleet of automobiles.
Keywords: power law process; repairable systems; missing cause of failure.
Anupap Somboonsavatdee is an Assistant Professor of Statistics in Department of Statistics and the Director of Center of Statistical Consulting and Research at Chulalongkorn Business School, Chulalongkorn University, Bangkok, Thailand. He received his PhD in Statistics from University of Michigan worked under advisory of Professor Vijay Nair and Professor Ananda Sen. His research interests include reliability, competing risks, incomplete data analysis, and data analytics on complex