题目：De-biasing the graphical Lasso in high-frequency data

主讲人：东京大学数学科学DB视讯(中国)Yuta Koike副教授

主持人：西南财经大学统计学院 常晋源教授

时间：2019年4月29日（星期一）15:00-16:00

地点：西南财经大学光华校区光华楼1007会议室

报告摘要：

Statistical inference for high-dimensional sparse inverse covariance matrices has been actively studied in the recent literature. A major approach for this issue is de-biasing a regularized estimator for the inverse covariance matrix. That is, we first construct an estimator for the inverse covariance matrix using some regularization to overcome the high-dimensionality, then we de-bias the estimator to recover the asymptotic normality which was lost due to regularization. In this talk we develop such a de-biasing procedure for the graphical Lasso in high-frequency data modeled as a discretely observed semimartingale. This allows us to implement statistical inference for the inverse quadratic covariation matrix in a high-dimensional setting.

主讲人简介：

Yuta Koike is an associate professor of statistics at University of Tokyo. In 2015, he got his Ph.D. in Mathematical Science from University of Tokyo. His current research interests include asymptotic statistics, financial econometrics, high frequency data, mathematical statistics, statistics for stochastic processes.