Interval Forecast for Smooth Transition Autoregressive Model

Abstract

In this paper, we propose a simple method for constructing interval forecast for smooth transition autoregressive (STAR) model. This interval forecast is based on bootstrapping the residual error of the estimated STAR model for each forecast horizon and computing various Akaike information criterion (AIC) function. This new interval forecast suggest definite and better coverage to the future sample path than the conventional method of using a multiple of standard error of the forecast distribution using bootstrap method. Simulation studies are used to illustrate the proposed method.

Keywords: AIC; bootstrap; Interval forecast; STAR