EcoSta 2021: Start Registration
View Submission - EcoSta2021
A0586
Title: Hybrid random weighting spectral test for multivariate white noise checking Authors:  Muyi Li - Xiamen University (China) [presenting]
Abstract: Vector autoregressive (VAR) models are one of the most popular tools in macroeconomic analysis. Hence the correct specification of the VAR models is crucial. To this end, we propose a frequency domain spectral test to check if the residuals from a fitted VAR model are multivariate white noises. The test statistics is a Cramer-von Mises (CM)-typed spectral test. The asymptotic null distribution can be obtained under mild conditions for more general unknown dependent structures on errors. In contrast to the time domain portmanteau tests, this spectral test is consistent and has nontrivial power against local alternatives by the order of the $\sqrt{n}$. Moreover, a blockwise hybrid random weighting method is employed to bootstrap critical values of the spectral CM test. The proposed bootstrapping procedure is easy to implement, and its first-order validity is justified. Monte Carlo simulation experiments and empirical data analysis are also reported.