报告题目：M-estimation for Jump-diffusion Models with Nonstationary High Frequency Data
报告人：王汉超（山东大学）
时间：2022.11.23（周三）14:00-15:00
腾讯会议：950-941-580
报告摘要：In this talk, we introduce the estimation for integrated jump- diffusion models with nonstationary high frequency data. As using the original data directly to estimate the unknown parameters in the models makes theoretical analysis almost not possible, we use a local approximation to derive desired estimation based on the original data. Moreover, as extreme jumps easily produce outliers, M-estimation procedure is considered for infinitesimal coefficients associated with integrated jump-diffusion models. However, the classical M-estimation can not possess the estimation consistency, the asymptotic normality and the rate of convergence, so we pro- pose a lag-1-based local M-estimation procedure to derive these important results. As a by-product, the designed conditions for these asymptotic results can even be weaker than those for classical independent identically distributed settings in the literature. The numerical studies suggest the ad- vantages of our method in bias reduction and computational efficiency. A real data example about the returns for stock index under five-minute high sampling frequency is analyzed for illustration.
王汉超，男，汉族，山东大学金融研究院教授，博士生导师。2011年于浙江大学数学系概率论与数理统计专业博士毕业。 主要从事概率统计极限理论及其应用的研究，特别在弱收敛，集中不等式与金融统计等领域发表了若干文章。与林正炎教授合作，在新加坡世界科技出版社出版专著Weak Convergence and Its Applications。与于志勇教授合作，在高等教育出版社出版教材《应用随机分析》。近几年来，在概率论，数理统计，计量经济等领域权威期刊上发表论文近二十篇。主持国家自然科学基金两项，作为骨干成员，参加国家重点研发计划项目两项。
