Testes para quebras estruturais em modelos não-lineares

Abstract

Since 1980 theoretical and empirical researches to detect changes in the parameters of non-linear models have been accumulated. In this context, there is still no consensus on the best statistical test for the detection of structural breaks. This work aims to make a study about main tests of the CUSUM family to detect points of parameters changes, and to compare their rejection rates. This behavior was investigated through Monte Carlo simulations in the family of Autoregressive Conditional Heteroskedasticity, ARCH, processes. The simulation indicates the robustness of each test on parameters change under the assumption of unconditional variance is little changed. Keywords: Points of changes, ARCH and CUSUM.