主讲人：中国科学技术大学 吴耀华教授
主 题：Random weighting approximation for Tobit response regression models with longitudinal data
主持人：统计学院 林华珍教授
时 间：2013年5月17日（星期五）上午：10：00-11：00
地 点：柳林校区通博楼B212
主 办：统计学院 科研处
主讲人简介：吴耀华，中国科学技术大学统计与金融系教授，博士生导师，现为中国数学会概率统计学会理事。一直从事数理统计的教学与研究工作，曾获国家级教学成果二等奖。从1986 年至今，在国内外学术刊物上发表论文二十余篇，主要研究领域包括：线性模型，非参数统计，回归分析和统计应用等方面。 应邀到第三世界科学院物理中心，香港中文大学、新加坡国立大学、美国哥伦比亚大学和加拿大皇后大学进行过合作研究。
讲座摘要：
Longitudinal data are widely studied and frequently applied in many fields such as medical studies and sociology. However, characters of studied subjects are often measured with a lower detection limit which results in left censored data. In this paper, we study Tobit response regression models with longitudinal data, and find that asymptotic distributions of the proposed quantile estimator and M-test statistic are related to some nuisance parameters including density function and the complicated correlation structure of error terms. To avoid estimating these nuisance parameters, randomly weighting method is employed to approximate to distributions of the studied statistics. We also build asymptotic properties of the proposed randomly weighting statistics and present a computation procedure for distribution approximation. Moreover, extensive simulations and a real data example are reported to show that the proposed method performs well in practical settings.
