講題:On dimension reduction of multivariate ARCH processes
講題摘要:
Many empirical time series such as asset returns exhibit the characteristic of time-varying conditional covariances, known as ARCH effects. Modeling multivariate ARCH processes, however, encounters several difficulties, including the curse of dimensionality. In this workshop, the speaker will review the methods used in the literature to reduce the dimension of multivariate ARCH processes, and introduce a new method developed by Hu and Tsay (2013) for analyzing a high-dimensional ARCH processes. This new method, called principal volatility component analysis (PVCA), can transform a k-dimensional ARCH processes to two parts: an r-dimensional series with ARCH effects and a (k-r)-dimensional series with no ARCH effects. An empirical analysis on the weekly log returns of 7 exchange rates against U.S. dollar from 2000 to 2011 shows that there exists a linear combination among the 7 exchange rates that have no ARCH effects. However, much work remains open for principal volatility component analysis, such as how to interpret the empirical findings in Economics and Finance and how to explore the properties of PVCA associated with non-zero eigenvalues.
To help audiences can get into the topic easily, the speaker will explain the motivation slowly from reviewing basic time series concept and analyzing a data set in the beginning of the workshop.