In this paper a robust analysis of the LM test for ARCH components is suggested.The main contribution is the introduction of the forward search method into the ARCH model family. Using very robust estimators of regression coefficients and graphicaldisplay of results, the effect of influential observations on estimates can be efficiently monitored. By means of both simulated and real financial series we show that identificationof the exact order of an ARCH model can be masked by the presence of few isolated big values. The presence of outliers does not affect the p−value of the test, while coefficientsand t−values are strongly influenced by few isolated extreme observations which lead to the identification of a wrong ARCH order.
Robust methods for the identification of ARCH components
GROSSI, Luigi;
2002-01-01
Abstract
In this paper a robust analysis of the LM test for ARCH components is suggested.The main contribution is the introduction of the forward search method into the ARCH model family. Using very robust estimators of regression coefficients and graphicaldisplay of results, the effect of influential observations on estimates can be efficiently monitored. By means of both simulated and real financial series we show that identificationof the exact order of an ARCH model can be masked by the presence of few isolated big values. The presence of outliers does not affect the p−value of the test, while coefficientsand t−values are strongly influenced by few isolated extreme observations which lead to the identification of a wrong ARCH order.
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
