报告题目：Group LASSO for Large Factor Models with Multiple Structural Breaks
报告人：涂云东副教授 北京大学
报告时间：2021年3月19日(周五)16:00-17:30
报告地点：中国科学院数学与系统科学研究院南楼N224教室（线下）， 腾讯会议 ID：439 1516 9050（线上）
Abstract：High dimensional factor models are becoming popular in modeling high dimensional time series in recent years. A common assumption of this type of model is that the factor loadings are time invariant. However, due to the systematic structure changes, the time instability of factor loadings attracts much research attention. It is widely recognized that ignoring the big structure changes in the factor loadings leads to the erroneous identification of an excessive number of factors. This paper reformulates the identification of structural breaks in factor loadings as a problem of detecting structural breaks in factor regressions, Group Fused Lasso based estimation procedures are proposed to identify the break dates. Our procedures are easy-to-implement, and overcome the obstacle of the classical methods which can only apply for detecting a single structure break, and extends easily to the scenario with multiple unknown breaks. The Monte Carlo simulation and real data demonstration illustrate that our procedures are fast implementable with desirable accuracy performance, and thus have practical merits.
涂云东:北京大学光华管理学院商务统计与经济计量系和北京大学统计科学中心联席副教授，研究员，入选首批“日出东方”光华青年人才。理论研究领域涵盖时间序列模型、非参数/半参数计量方法、模型选择和模型平均、网络数据建模、金融计量、信息计量经济学、模型设定检验等；应用研究包含宏观经济预测、价格指数建模、金融市场预测、环境污染预测、新冠肺炎预测等。研究成果发表于Journal of Econometrics, Econometric Reviews, Journal of Business and Economic Statistics, Oxford Bulletin of Economics and Statistics，Statistica Sinica，Journal of Empirical Finance，Computational Statistics and Data Analysis等多个知名专业期刊。
