Information Criteria for the Identification of the Box-Cox Transformation Model

摘要

This paper deals with the information criteria applied t.o the identification of nonlinear model which with Box-Cox transformation (BCT). For the purpose of estimating the BCT parameter and the clan of regressions, general information criterion (GIC) is proposed. It is seen that Akaike's information criterion (AIC) for the identification of the BCT model is a special case of the developed GIC. To compare the power performance of the GIC and the AIC. Monte Carlo simulations are conducted. It shows that as for the estimate of the BCT parameter, the GIC is a little precise than the AIC; but it is on the contrary in order identification for the BCT polynomial regression model.