A SETAR model for Canadian GDP: non-linearities and forecast comparisons

Abstract

This paper investigates the forecasting performance of the non-linear time series SETAR model by using Canadian GDP data from 1965 to 2000. Besides the within-sample fit, the forecasting performance of a standard linear ARIMA model for the same sample has also been generated for comparative purposes. Two forecasting methods, one-step-ahead and multi-step-ahead forecasting, are compared for each type of model.