报告题目：An automatic MDDM-based test for martingale difference hypothesis

报告时间：2024年4月17日 上午10:00

会议链接：https://meeting.tencent.com/dm/sIGbbpsRCBFz

会议 ID：844-739-415

主办单位：数学与统计学院/科研处

报告人：王国长

报告人简介：王国长，现任暨南大学经济学院统计学系教授、博士生导师，中组部青年拔尖人才支持计划获得者。2012年取得统计学博士学位，2012-2014年在中国科学院应用所从事博士后研究工作，2017-2018年赴香港大学统计与精算系学术访问1年。主要研究方向为函数型数据分析、时间序列、充分性降维等，迄今为止在JoE, JBES, Sinicas等重要学术期刊接收和发表论文20余篇。主持国家级项目4项，省部级项目3项。任中国现场统计研究会资源与环境统计分会常务理事；广东省现场统计协会常务理事，秘书长。

摘要：To check the error term whether is a marginal difference sequence (MDS) in the multivariate time series model with parametric conditional mean is a very important problem. Since if the error term is a MDS, which indicates that the proposed model is right, if not, which means that there is a lack of fit in the postulated conditional mean specification and can lead to misleading statistical inferences and suboptimal point forecasts, resulting in erroneous conclusions. The test based on martingale difference divergence matrix (MDDM) is an useful statistical method to test the MDS in the multivariate time series model, but the MDDM-based tests should specify the number of the lag. To solve this problem, we propose a data-driven MDDM-based test to select the lag automatically. This test has two advantages over existing tests: firstly, the researcher does not need to specify the lag, since the test automatically chooses this number from the data; secondly, under the null hypothesis, the lag is equal to one and which can save a lot of computational costing. Finally, we use numerical studies including simulation and real data analysis to illustrate the usefulness of the proposed test.