．單位首頁
．系所簡介
．系主任
．強力師資
．研究所
．課程概況
．各類學程
．期刊論文
．招生資訊
．演講專區
．榮譽榜
．系學會
．系友天地
．就業訊息
97年3月28日(五) 10:00 ~11:00 A.M.
演講者姓名： Dr. Richard Gerlach
演講者服務單位： Discipline Econometrics and Business Statistics, University of Sydney, Australia
Testing for Bivariate Asymmetry Using Approximate Bayesian Model Selection
Abstract
A Bayesian model selection procedure is proposed to simultaneously: (i) test for asymmetry in bivariate financial time series; (ii) identify an appropriate heteroskedastic model from a family of bivariate GARCH models. This model family captures the major stylised features of financial stock returns: fat tails, excess kurtosis, volatility clustering and volatility, covariance and mean asymmetry in a parsimonious specification with easily identifiable and interpretable parameters. Both the necessary and sufficient conditions for stationarity and positive definiteness are enforced via a specific diffuse prior distribution. An adaptive Markov chain Monte Carlo approach is proposed for estimation and inference, and combined with a computationally efficient, but approximate, formal model selection approach. A simulation study illustrates favourable estimation properties and strong model selection performance, indicating that an accurate approximation has been obtained, for practical purposes. An empirical study on stock returns from international markets is presented and posterior model probabilities are estimated, in each case revealing clear volatility and covariance asymmetry across markets.
國科會數學推動中心補助
