报告人：刘志
报告地点：数学与统计学院415室
报告时间：2019年06月26日星期三14:00-15:00
邀请人：
报告摘要：
The statistical inference of volatility matrix suffers from the effects of variate features of ultra-high-frequency data, such as market microstructure noise, asynchronous trading, multiple records, etc. In particular, the tick-by-tick transaction records are typically asynchronous among different assets and multiple records frequently appear. In this talk, I will review the existing feasible approaches. These methods, however, have to discard a large part of data, to keep the estimators consistent. Then I will introduce a data-efficient estimator of the realized volatility matrix which utilizes all of the data. The estimator is created by combining the congregation of the data within the synchronized time intervals and pre-averaging smoothing methods. We have established the asymptotic normality of proposed estimator. A studentized version of the central limit theorem is also proposed. I will also show improvement in estimation efficiency. Through a variety of synthetic data experiments, we assess the finite sample performance of proposed estimator and make comparison with existing methods. Finally, we implement the estimator with real NASDAQ high-frequency financial data sets.
主讲人简介：
刘志，澳门大学数学系副教授。主要研究方向包括: 金融超高频数据分析、随机过程统计推断，等。其研究近年来获得了澳门政府以及国家自然科学基金等多项基金的资助，在统计学、金融和生物信息国际期刊发表论文40余篇，主要研究成果发表在AoS、JASA、JoE、JBES、Bioinformatics、SPA、ET等相关研究方向的权威期刊上。
