Least Squares Predictors for Threshold Models: Properties and Forecast Evaluation

Abstract

The forecasts generation from SETARMA models is presented and discussed. In particular, least squares, plug-in and combined predictors are pointed out even highlighting the main problems that arise when forecasts have to be computed for this class of threshold models. An empirical example to three stock market indexes shows the performance of the proposed predictors giving evidence in favor of the forecasts combination.