A Goodness of Fit Test for Threshold GARCH Models

Abstract

In this paper, an entropy based goodness of fit test is applied to conditional heteroscedasticity models such as the threshold GARCH model. (cf. Lee et al. (2011)). The asymptotic distribution of the test statistic is derived under the null hypothesis so that the test can be taken into practice. In this paper, a bootstrap method is applied in the simulation study to deal with the simple vs. simple test and the composite hypothesis case is stated theoretically. In the simulation study, the performance of the test for four pairs of different , values and model coefficients are evaluated through Monte Carlo simulations. The result shows an adequate performance of the test. Finally, a real data analysis is conducted by using the stock price of Intel Corporation from January 2009 to November 2012 with 988 observations.