A0797
Title: Information matrix tests for switching regression models
Authors: Enrique Sentana - CEMFI (Spain) [presenting]
Dante Amengual - CEMFI (Spain)
Gabriele Fiorentini - University of Florence (Italy)
Abstract: The EM principle implies that the moments underlying the information matrix test for multivariate switching regression models are the filtered values of the moments that the information matrix test would test if one knew the latent component each observation belongs to. Thus, components are identified related to the conditional heteroskedasticity, conditional and unconditional skewness, and kurtosis of the multivariate regression residuals for each of the regimes. The Monte Carlo simulations indicate that the expressions obtained by numerical integration for the asymptotic covariance matrix of those empirical moments adjusted for sampling variability in the maximum likelihood parameter estimators provide reliable finite sample sizes and good power against various alternatives, especially combined with the parametric bootstrap.
