Title: Random weighting the Portmanteau tests for multivariate white noise with unknown dependent structure
Authors: Muyi Li - Xiamen University (China) [presenting]
Abstract: The Ljung-Box portmanteau test is one of the most popular model diagnostic tools. However, when the errors are not i.i.d.random variables, the classical portmanteau test does not follow asymptotically chi-squared distribution. We employ the random weighting method to bootstrap the critical values of Ljung-Box portmanteau test in multivariate time series models with unknown dependent white noise. A set of Monte Carlo experiments demonstrate the practical relevance of this method. Real examples on the USD-MYR and USD-SGD five-day exchange rates illustrates the merits of our testing procedures.