主 题：Estimating Spatial Autocorrelation with Sampled Network Data
主讲人：王汉生教授
主持人：林华珍教授
时 间：2014年5月7日下午4:40-5:40
地 点：通博楼B座212学术会议室
主办单位：统计学院 统计研究中心 科研处
主讲人简介：
北京大学光华管理学院商务统计与经济计量系,嘉茂荣聘教授，博导，系主任。北京大学商务智能研究中心、主任。博雅立方科技有限公司首席科学家。微信公众号“狗熊会”创始人。1998年北京大学数学学院概率统计系本科毕业，2001年美国威斯康星大学麦迪逊分校统计系博士毕业。2003年加入光华至今。国内外各种专业杂志上发表文章七十余篇，并（合）著有中英文专著各一本。国际统计协会、英国皇家统计协会、美国统计协会（、美国数理统计协会、泛华国际统计协会的会员。先后历任以下国际学术刊物副主编（Associate Editor）：The Annals of Statistics (2008—2009), Computational Statistics & Data Analysis (2008—2012)，Statistics and its Interface (2010—现在), Journal of the American Statistical Association (2011—现在)，以及Statistica Sinica (2011—现在)。Journal of Business and Economics Statistics (2012—现在), Science China: Mathematics (2013—现在）
内容提要：
Spatial autocorrelation is a parameter of importance for network data analysis.To estimate spatial autocorrelation, maximum likelihood has been popularly used. However, its rigorous implementation requires the whole network to be observed. This is practically infeasible if network size is huge (e.g., Facebook, Twitter, Weibo, WeChat, etc). In that case, one has to rely on sampled network data to infer about spatial autocorrelation.By doing so, network relationships(i.e., edges) involving unsampled nodes are overlooked. This leads to distorted network structure and underestimated spatial autocorrelation.To solve the problem, we propose here a novel solution. It makes use of the fact that spatial autocorrelation is typically small. This enables us to approximate the targeting likelihood by its first order Taylor's expansion. Depending on the choice of the likelihood,we obtain respectively an approximate maximum likelihood estimator (AMLE)and paired maximum likelihood estimator (PMLE). We show theoretically that both methods are consistent and asymptotically normal with identical asymptotic efficiency. However, the difference is that PMLE is
