報告題目：Recent Advances in Hidden Markov Models: Inferences and Applications
報 告 人：宋心遠 香港中文大學 教授
報告時間：2021年4月2日 下午 13:30-14:30
報告地點：騰訊會議
點擊鍊接入會，或添加至會議列表：
https://meeting.tencent.com/s/pO4QmqI4Zn6x
會議 ID：540 245 947
校内聯系人：趙世舜 zhaoss@jlu.edu.cn
報告摘要：
This talk presents recent advances in hidden Markov models (HMMs) for longitudinal data analysis. HMMs are commonly used to simultaneously investigate longitudinal observation process and the underlying dynamic transition process. The statistical inferences and applications of advanced HMMs are discussed. First, we develop a Bayesian adaptive group lasso procedure to conduct variable and function selection in the context of semiparametric HMMs. A basis expansion is used to approximate the nonparametric functions. Multivariate conditional Laplace priors are introduced to facilitate adaptive penalization on regression coefficients and various groups of basis expansions. An efficient Markov chain Monte Carlo algorithm is developed to identify important covariate and functional effects in the conditional and transition models. The proposed model is applied to the Alzheimer's Disease Neuroimaging Initiative study. Moreover, we discuss several recent developments on HMMs and possible extensions, such as the order selection of HMMs, varying-coefficient HMMs with zero-effect regions, quantile HMMs, and joint analysis of HMMs and survival models with longitudinal and time-to-event data.
報告人簡介：
宋心遠教授，香港中文大學統計系主任。宋心遠教授的研究方向是潛變量模型，貝葉斯方法，統計計算和生存分析等。同時還擔任多個國際期刊包括《Psychometrika》，《Biometrics》，《Computational Statistics & Data Analysis》和《Structural Equation Modeling: A Multidisciplinary Journal》的副主編或編委。已在國際期刊發表超過100篇論文，近期論文主要發表于《Journal of the American Statistical Association》，《Biometrika》，《Biometrics》，《Bioinformatics》，《Psychometrika》，《Quantitative Finance》等期刊。
