Events at CERGE-EI

Pavel Čížek, Ph.D.

Tilburg University, The Netherlands

Authors: P. Čížek and C.-H. Koo

Abstract: A new semiparametric time series model is introduced – the semiparametric transition model – that generalizes the threshold and smooth transition models by letting the transition function to be of an unknown form. The estimation strategy is based on alternating the conditional and unconditional least squares estimation of the transition function and the regression parameters, respectively. The consistency and asymptotic distribution for the regression-coefficient estimator of the semiparametric transition model are derived and shown to be first-order asymptotically independent of the nonparametric transition-function estimates. Monte Carlo simulations demonstrate that the estimation of the semiparametric transition model is more robust to the type of transition between models than the parametric estimators of the threshold and smooth transition models.

JEL codes: C14, C21, C22

Keywords: local linear estimation, nonlinear time series, semiparametric estimation, regime-switching models

Full Text:  “Semiparametric Transition Models”