讲座时间：2022年5月6日 14：00-15：30

讲座地点: 腾讯会议 673 748 025

主讲人：Jun Yi PENG ZHOU

讲座题目：Self-Normalized Stationary Tests

讲座简介：In this paper, we propose an extension for testing stationarity without the need for a consistent estimate of the long-run variance (LRV).  The statistics are a transformation of the KPSS test proposed by Kwiatkowski, Phillips, Schmidt, and Shin (1992),  the modified R/S statistic by Lo (1991), and the V/S statistic by Giraitis,  Kokoszka, Leipus and Teyssiere (2003). Under persistent autocorrelation, the LRV estimator proposed by Newey and West (1897,1994) and Andrews (1991) may not be reliable and introduces size and power distortions on the tests. Also, the practitioner has to choose the kernel truncation lag that is ultimately arbitrary. To improve the performance of the tests for stationarity and make it robust, we propose the use of self-normalizing methods, for example,  the range-based self-normalization by Hong, McCabe and Sun (2020), and Fixed-B asymptotic by Kiefer and Vongelsang (2005), and Amsler, Schmidt, and Vogelsang (2009) to control the effect of the LRV. The self-normalized tests are not consistent in the sense that under the alternative hypothesis does not diverge as the sample size increases. To recover the test's consistency, we devise a mechanism similar to the power enhancement mechanism proposed by Fan, Liao, and Yao (2015). Under the null hypothesis, this mechanism is asymptotically negligible. However, under the alternative hypothesis, the mechanism diverges as the sample size increases. We show that this mechanism also enhances the power of the self-normalized stationarity tests.

主讲人介绍：Dr. PENG Jun Yi is currently an assistant professor at the Institute for Advanced Economic Research at Dongbei University of Finance and Economics. He received his Ph.D. degree from Universidad Carlos III de Madrid in 2019. His research interests are Econometric Theory, Time Series Modelling, and Finance. His current research studies threshold models with unit roots from two different perspectives, a univariate approach by the stochastic unit root models where the randomness of the unit root is driven by other economic variables. On the other hand, from a multivariate perspective by introducing threshold effects in the cointegration relation allowing for the presence of multiple equilibria.