|
Abstract:
|
In this talk, we propose a empirical likelihood based model-building for GARCH models. We construct the time domain empirical likelihood, which computes GARCH parameters and tests for autocorrelation of the estimated residuals simultaneously. The test statistics can be also used for order determination of GARCH models. The estimators of the GARCH parameters have the same asymptotic variance as the conditional maximum likelihood estimators. Test statistics have asymptotic chi-squared distributions. Results of some simulation studies are also reported.
|
