Threshold Autoregressions

Abstract

The previous chapter introduced a number of nonlinear (univariate) models, which had in common the idea of smooth transition between regimes. The most popular class of such models is the smooth transition autoregressive — or STAR — class, of which the exponential and logistic members are the most frequent in application, giving rise to the acronyms ESTAR and LSTAR. Another newer class of models, also considered in the previous chapter, motivated by similar considerations, was the bounded random walk, BRW.