A0602
Title: Block wild bootstrap based Ljung-Box test for VAR model with time-varying variance
Authors: Taewook Lee - Hankuk University of Foreign Studies (Korea, South)
Younjae Lee - Hankuk University of Foreign Studies (Korea, South) [presenting]
Abstract: The goodness-of-fit test of vector autoregressive (VAR) models is investigated with time-varying variance. While the conventional Ljung-Box (LB) test is assumed to follow the chi-square distribution asymptotically, it has been found that the LB test can deviate significantly from the chi-square distribution for VAR models with time-varying variance. The block wild bootstrap-based (BWB) LB test is proposed, and it has been proved that our BWB-LB test achieves a correct critical region for the goodness-of-fit test of VAR models with time-varying variance. The simulation study shows that the BWB-LB test achieves the correct sizes and comparable powers in finite samples.
