A Joint Model for Multiple Longitudinal Data with Different Missing Data Patterns and with Applications to HIV Prevention Trials

报告人：吴婧副教授（University of Rhode Island）

报告时间：2025年6月15日，上午10：30-11：10

地点：维格堂113

摘要

In longitudinal clinical trials, it is common that mixed types of outcomes are collected

on the same subject over time. It is also routinelyen countered that all outcomes may

be subject to substantial missing values due to dropout and intermittent missingness.

Additionally, the missing data patterns of the mixedtypes of outcomes are usually the

same for dropout while different for intermittent missingness. In this paper, a sequential multinomial model is adopted for dropout and subsequently, a new joint conditional model is constructed for intermittent missingness of mixed types of outcomes. The new model captures the complex structure of missingness and incorporates dropout and different intermittent missingness simultaneously. Two types of outcomes (binary and count) are considered in this paper. A mixed-effects probit regression model and a zero-inflated Poisson mixed-effects regression model are assumed for the longitudinal binary and count response data, respectively. We further show that the joint posterior distribution is improper if uniform priors are specified for ther egression coefficients under the proposed model. An efficientGibbs sampling algorithm is developed using a hierarchical centering technique. A modified logarithm of the pseudo marginal likelihood (LPML) and a new concordance measure criterion are used to compare the models under different missing data mechanisms. An extensive simulation study is conducted to investigate the empirical performance of the proposed methods, andthe methods are further illustrated using real data from an HIV preventionclinical trial.

个人简介：

Jing Wu is an associate professor anddirector of graduate studies in statistics at the University of Rhode Island. She received her BS in mathematics from Shanghai Jiao Tong University and her Ph.D. in statistics from the University of Connecticut. Her research interests primarily lie in missing data, big data, Bayesian statistics,survival and longitudinal data. She is currently an associate editor of Statistics and Its Interface and production editor of the New England Journalof Statistics in Data Science. Dr. Wu has published research articles in Journal of Computational and Graphical Statistics, Statistics in Medicine,Statistica Sinica, Technometrics, PNAS, JAMA Oncology, and Journal of Clinical Oncology, and so on.

邀请人：刘芳