Masters Oral Defense: "A Simulation Study of Bipower and Tresholded Realized Variations for High-Frequency Data"

Zheng Xu, Washington University in Saint Louis

Abstract: In the framework of general time-continuous stochastic model, multipower variation and threshold type variation are two main trendy types of nonparametric estimators with a wide range of applications. Given the high-frequency data, these methods aim to estimate the integrated volatility of the diffusion component in Ito type stochastic process. In this paper, we will focus on the bipower variation (BPV) and thresholded realized variations (TRV) for some classical Levy stochastic models. We use R to construct Merton jump diffusion model, Kou jump diffusion model, variance gamma model and normal inverse Gaussian model and test the performance of estimators in each model with different time lags based on the high-frequency data. We also introduce some types of threshold sequences that widely applied in the TRV estimator. Simulation results demonstrate that the TRV estimator with an optimal threshold sequence presents a better performance on the efficiency of jump detection and the accuracy of integrated volatility estimation in each model.

Hosts: Nan Lin & Jose Figueroa-Lopez