实验研究方法论坛：生物统计与实验研究前沿

时间：2023年4月21日 13:30-17:30 

地点：综合楼644会议室

【报告一】Phase-Type Distributions for Lifetime Data Analysis

报告人：王昱栋
报告时间：13：30-14：30
报告人简介：王昱栋，新加坡国立大学工业系统工程与管理系博士后研究员。其主要研究方向为生存分析，可靠性工程和应用统计。其研究成果发表在Journal of the Royal Statistical Society Series B、Biometrics、IEEE Transactions on Reliability等期刊.
报告摘要：Product return data, such as warranty claims, are usually subject to two layers of right censoring. The first layer, called warranty censoring, applies to the product lifetime due to a fixed warranty limit. The second layer, called end-of-study censoring, applies to the sum of the sales lag and the lifetime due to the end-of-study date for the data collection. The two-layer censoring in the product return 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 to compute the maximum likelihood estimators. Comprehensive simulations are conducted to evaluate the performance of the proposed method and compare it with the inverse probability of censoring weighted approach, which is applicable in the absence of warranty censoring. Two real examples from production-delivery supply chains are analyzed by the proposed method to provide guidance on warranty reserve management for the manufacturers.

报告人：王平平
报告时间：14：30-15：30
报告人简介：王平平，博士毕业于华东师范大学统计学专业，现就职于南京财经大学经济学院统计系。主持国家自然科学基金青年项目、江苏省自然科学基金青年项目、教育部人文社会科学基金青年项目各一项。研究方向为可靠性统计、空间统计，论文发表在IEEE Transactions on Reliability, RESS, statistics in medicine等期刊上.
报告摘要：Spatial ordinal data observed separately for multiple subjects are common in biomedical research, yet statistical methodology for such ordinal data analysis is limited. The existing methodology often assumes a single realization of spatial ordinal data without replications, a commonplace in disease mapping studies, and thus are not directly applicable. Motivated by a dataset evaluating periodontal disease (PD) status, we propose a multisubject spatial ordinal model that assumes a geostatistical spatial structure within a regression framework through an elegant latent variable representation. For achieving computational scalability within a classical inferential framework, we develop a maximum composite likelihood method for parameter estimation, and establish the asymptotic properties of the parameter estimates. Another major contribution is the development of model diagnostic measures for our dependent data scenario using generalized surrogate residuals. A simulation study suggests sound finite sample properties of the proposed methods. We also illustrate our proposed methodology via application to the motivating PD dataset. A companion R package clordr is available for easy implementation.

报告人：周世荣
报告时间：15：30-16：30
报告人简介：周世荣，博士毕业于华东师范大学统计学专业，现就职于温州大学数理学院统计与信息科学系. 研究方向为可靠性统计、贝叶斯统计，论文发表在IEEE Transactions on Reliability, Reliability Engineering & System Safety等期刊上.

报告摘要：In the field of reliability engineering, the gamma process plays an important role in modeling degradation processes. However, extracting lifetime information from product degradation observations has long been suffering from both ineffective modeling techniques and inefficient statistical inference methods. To overcome these challenges, we propose a reparameterized gamma process with random effects in this paper. Compared with the classical gamma process, the proposed model has a more intuitive physical interpretation. In addition, statistical inference for the model can be readily done through the variational Bayesian algorithm. Combining with the Gauss-Hermite quadrature and the Laplace approximation, the algorithm yields closed-form variational posteriors for the proposed model. Its superiority over two other inference methods (expectation maximization and Monte Carlo Markov Chain) in terms of computational efficiency and estimation accuracy is demonstrated by simulation.

报告人：汤银才
报告时间：16：30-17：30
报告人简介：汤银才，华东师范大学教授，博士生导师，《Statsitical Theory and Related Fields》执行主编，《应用概率统计》和《华东师范大学学报》编委,中国运筹学会可靠性分会常务理事、中国现场统计研究会可靠性工程分会副理事长，中国现场统计研究会大数据统计分会常务理事、副秘书长，中国数学会概率统计学会、中国现场统计研究会计算统计分会、上海市工业与应用数学会理事，主持并完成国家自然科学基金3项，其他各类项目20多项，在国内外学术刊物上发表论文100多篇，先后获得华东师范大学研究生教育优秀教师奖，上海市科学技术三等奖，上海市教育发展基金会申银万国奖，华东师范大学优秀任课教师奖，上海市教学成果三等奖，上海市科技进步三等奖，全国统计科学技术进步二等奖，上海市统计科学研究成果课题类一等奖等荣誉。著有《R语言与统计分析》、《可靠性统计》、《贝叶斯统计》。

报告摘要：In this paper, we propose a PG-INLA algorithm which is tailored to the unidimensional and multidimensional 2-PL IRT models. The proposed PG-INLA algorithm excute a computationally efficient data augmentation strategy via the PolyaGamma variables which could avoid the low computational efficiency for the IRT models with logistic link functions. Meanwhile, combined with the advanced and fast INLA algorithm, the PG-INLA algorithm is accurate and computationally efficient. In addition, we provide details on the derivation of posterior and conditional distributions of IRT models, the method of introducing the Polya-Gamma variable into Gibbs sampling, and the implementation of the PG-INLA algorithm for both unidimensional and multidimensional cases. Through simulation studies and an application to the analysis of the IPIP-NEO personality inventory, the performance of the PG-INLA algorithm is assessed. Extensions of the proposed PG-INLA algorithm to other IRT models are also discussed.