A0629
Title: Bootstrapping robust goodness-of-fit tests for GARCH models
Authors: Muyi Li - Xiamen University (China) [presenting]
Abstract: A random weighting (RW) bootstrap approach is considered to conduct the robust goodness-of-fit tests for the generalized autoregressive conditional heteroskedastic (GARCH) models, which is estimated by the least absolute deviation (LAD) method. The RW bootstrap method perturbs both the minimand of the objective function and autocovariances of the transformed residual sequence, such that the test is applicable for very heavy-tailed innovations with only finite fractional moments. The testing procedure is easy to implement and the test statistic is robust to the choice of the random weights. The first-order consistency of the RW procedure is proved and the finite sample performance of the proposed test is assessed by numerical experiments. Finally, a real data analysis illustrates the usefulness of the test.
