统计与管理学院2015年学术报告第20期
【主 题】 Empirical Process Based Specification Testing for Ergodic Diffusions Sampled From High Frequency Data
【报告人】 陈 强 博士
上海财经大学经济学院
【时 间】 2015年5月18日（星期一）16:00－17:00
【地 点】 上海财经大学统计与管理学院大楼1208室
【语 言】 英文
【摘 要】 This paper develops two joint tests for the parametric drift and volatility function of a diffusion model based on empirical processes and the Khmaladze (1981)'s martingale transformation. The tests are in the spirit of classical Kolmogorov-Smirnov and Cramer-von Mises tests, and are shown to be asymptotically distribution-free. They are consistent against a large class of fixed alternatives under different schemes, and have nontrivial power against a class of different rate local alternatives for drift and volatility. Monte Carlo simulations show that the tests perform quite well in finite samples, and our tests outperform the nonparametric test of Hong and Li (2005). The proposed tests are then applied to testing short-term interest rate models, using data of Treasury bill rate and Eurodollar deposit rate. The empirical results show that, the correct specification of volatility function is much more important than that of drift function, and it is harder to detect drift misspecification than to detect volatility misspecification, which is consistent with the theory.
