Abstract:
His paper proposes a nonparametric test for conditional independence (CI) by estimating the conditional density ratio (CDR) function, i.e. the ratio of the product of two conditional probability densities to their joint conditional density, through a simple weighted nonparametric least-squares regression (WNPLSE). The proposed method owns several superiorities over existing alternatives: first, the test statistic is invariant to the monotone transform of the data distributions and has a closed-form expression that enhances the speed and efficiency in computation; second, the CDR estimator satisfies empirical moment restrictions such that extreme values are unlikely to be obtained; third, the proposed test is consistent against all departures. The key driver is that the CDR is a constant under the null hypothesis so that its approximation bias is eliminated. A small-scale simulation study indicates that the proposed test outperforms the existing alternatives in various dependence structures.
About the Speaker:
Dr. Chunrong Ai is Professor of Economics and Presidential Chair Professor in the School of Management and Economics, Chinese University of Hong Kong, Shenzhen. He is the Director of the Social and Behavioral Big Data Lab at The Shenzhen Research Institute of Big Data and Associate Director of Shenzhen Finance Institute. Before that, he was Professor of Economics and Florida Term Professor at the University of Florida. He was the founding Dean of the Institute of Statistics and Big Data, Renmin University of China, and the founding Dean of the School of Statistics and Management, Senior Associate Dean of the Institute for Advanced Research and Associate Dean of Business School, Shanghai University of Finance and Economics.
Dr. Ai’s primary research interests are in the area of econometrics, empirical industrial organization, empirical finance, and Chinese economy. Recently he focused on the interdisciplinary research between data science and behavioral big data. His research outputs are published in the reputable scholarly journals.
Your participation is warmly welcomed!
欢迎扫码关注北大统计科学中心公众号，了解更多讲座信息!
