Generalized Standardized-Residuals-Based Correlation Tests for GARCH-type Models
2005/03/01
研討會日期 : 2005-03-01
時間 : 15:00
主講人 : Prof. Chen Yi-Ting
地點 : B棟110室
演講者簡介 : Prof. Chen Yi-Ting 為 Ph.D. in Economics, National Taiwan University (1998)。
其主要研究領域為Econometrics Theory、Time Series Analysis及Empirical Finance。
現為本院人文 社會科學研究中心副研究員。
演講摘要 : Motivated by the potential size and power problems of the conventional autocorrelation tests, in this paper we propose a flexible class of standardized-residuals-based correlation tests. By correcting the estimation error effect, the proposed tests are applicable for a variety of time series (or dynamic) models in a general conditional heteroskedasticity context. Moreover, their power directions can be easily extended by considering different autocorrelations and cross-correlations. By the generality of the proposed method, we demonstrate the applicability and limitation of the Box-Pierce, McLeod-Li, and Li-Mak tests. We also discuss a class of power-transformation correlations in detail, and illustrate that the resulting tests should be especially useful for discovering serial correlation, regime-switching correlation, volatility clustering, and volatility asymmetry and testing the time series models with the AR, SETAR, GARCH, EGARCH, or other similar specifications. The Monte Carlo simulation supports the validity of the proposed method and the usefulness of the correlation tests in dependence structures exploration and model diagnostic check. Finally, we apply the proposed tests to an empirical study of financial time series.
