Численное сравнение V-MLR и CUSUM методов обнаружения структурных сдвигов для кусочно-заданных GARCH-моделей
In this article we propose a new method of structural breaks detection for GARCHmodels called V-MLR. We use two numerical experiments consisting of 10 000 simulations to compare our V-MLR method with the well-known CUSUM method. In the first experiment with a single structural break V-MLR method found the correct number of structural breaks in 91% of cases, and CUSUM — in 85% of cases. The accuracy of the structural break detection by both methods was proved to be comparable. In the second experiment without any structural breaks the V-MLR method indicated the absence of structural breaks in 99% of cases, while CUSUM — only in 91% of cases. Thus, the numerical experiments suggest that V-MLR and CUSUM methods have comparable accuracy of structural breaks moment detection, but the proposed V-MLR method has a greater sensitivity to structural breaks compared to CUSUM.