报告题目：Recent Developments in Observation-driven Count Models
报告人：  美国陶森大学 (Towson University) 崔云炜 博士
报告时间：2017年6月7日（周三）下午 15:00-16:00
报告地点：行政楼601会议室

 

报告摘要：

Integer-valued time series models have attracted considerable interest in recently years. In many fields integer-valued time series arise naturally in the form of counts. One line of research on count time series employs generalized linear models, where serial dependence among the observed counts is introduced through a state process. Previous works focus on Poisson deviates, i.e., conditional on the information up to date the observed count follows a Poisson distribution. Instead of specifying a particular type of deviates, we assume that the deviates come from the one-parameter exponential family. Our setting unifies Poisson, binomial, and negative binomial deviates in a common framework. The probability stability conditions such as stationarity and ergodicity of the proposed models are investigated. The statistical inference is conducted based on the conditional maximum likelihood estimators (CMLE). Under certain regularity conditions, the strong consistency and asymptotic normality of CMLE are established. INGARCH(p,q) models with negative binomial deviates are verified to satisfy the regularity conditions and thus the main results are applicable.

 

报告人简介：

    崔云炜博士，男，美国陶森大学数学系助理教授。本科就读于天津大学，后赴美国Clemson大学攻读硕士和博士学位，2009年获得博士学位后，先后在美国休斯敦大学（University of Houston Downtown）和美国陶森大学（Towson University）任职。崔云炜博士在时间序列方向有着突出的建树，在国际著名的期刊《Biometrika》、《Journal of Statistical Planning and Inference》、《Journal of Time Series Analyisis》、《Statisticas & Probability Letters》等发表论文多篇，多次在重要的国际会议和workshop中进行学术报告，担任多个期刊的评审工作。

欢迎老师和同学们参与讨论！