Smooth buffered autoregressive time series models

Abstract

In this paper, we study a new model called the smooth buffered autoregressive (SBA) model. A sufficient condition is given for geometric ergodicity. A conditional least squares (CLS) estimation procedure is discussed, and consistency and normality of the estimators are derived. We investigate the effectiveness of our methods by simulation studies. Two applications are considered: annual sunspot number and the U.S. unemployment rate.