Survival data, such as warranty claims and disease registry data, are usually subject to two layers of right censoring. The first layer, which is called lifetime censoring, applies to the lifetime due to a fixed warranty limit or a competing risk. The second layer, which is called end-of-study censoring, applies to the sum of the sales lag (or reporting delay) and the lifetime due to the end-of-study date for the data collection. An unreported subject would either have a lifetime longer than the lifetime censoring limit or the sum of the sales lag (or reporting delay) and the lifetime longer than the end-of-study date. The two-layer censoring in the lifetime data renders traditional nonparametric methods for right-censored data inapplicable. This study develops a generic method for the two-layer censored data using acyclic phase-type distributions (APHDs) in the canonical form. The APHD estimators can be regarded as nonparametric sieve estimators since the family of APHDs is dense in the field of all positive-valued distributions. Based on the property that the class of APHDs is closed under convolution, a dedicated expectation-maximization algorithm is proposed for parameter estimation. Comprehensive simulations are conducted to evaluate the performance and compare with the inverse probability of censoring weighted approach, which is applicable in the absence of lifetime censoring. Two real examples are used to illustrate the proposed method.

Biography:

Dr Ye received double bachelor's degrees in Materials Science and Engineering, and Economics from Tsinghua University in 2008 and he obtained his Ph.D. from the National University of Singapore (NUS). He is currently an Associate Professor and Dean’s Chair in the Department of Industrial Systems Engineering and Management, NUS. His research focuses on resilience and reliability engineering, industrial statistics, and data-driven operations management. He has published more than 100 papers in leading journals in reliability, statistics, and operations management, including Bernoulli, Biometrics, Biometrika, JASA, JMLR, JRSS-B, JRSS-C, Technometrics, JQT, IISE Transactions, OR, MSOM, POMS, and the IEEE Transactions series.

In this talk, I will discuss a novel multivariate point process regression model for a large-scale physically distributed network infrastructure with two failure modes, i.e., primary failures caused by the long-term usage and degradation of the asset, and cascading failures triggered by primary failures in a short period. Field pipe failure data from a UK-based water utility are exploited to support the rationale of considering the two failure modes. The two failure modes are not self-revealed in the field data. To make the inference of the large-scale problem possible, a time window for cascading failures is introduced, based on which the likelihood of the pipe failure process can be decomposed into two parts, one for the primary failures and the other for the cascading failure processes modulated by the primary failure processes. The window length for cascading failures is treated as a tuning parameter and it is determined through maximizing the likelihood based on all failure data. To illustrate the effectiveness of the proposed model, two case studies are presented based on real data from the UK-based water utility. Interesting features of the cascading failures are identified from massive field pipe failure data. The results provide insights on more advanced modelling and practical decision-making for both researchers and practitioners.

Biography:

Dr. Xun Xiao is currently a Lecturer in Statistics at the Dept. of Mathematics and Statistics, University of Otago, New Zealand. He received B.Sc. in Statistics from the University of Science and Technology of China in 2011 and Ph.D. degree from the Dept. of Systems Engineering and Engineering Management at City University of Hong Kong in 2016.  His current research focuses on industrial statistics and point process modelling.