Threshold variable selection by wavelets in open-loop threshold autoregressive models (Q1962219)

The authors deal with identification of the general open-loop threshold autoregressive models TARSO (the classical SETAR models are a special case). The time delay and thresholds are detected and estimated by checking the variation of the empirical wavelet coefficients. These coefficients are defined using results by \textit{Y.K. Truong} and \textit{C.J. Stone} [Ann. Stat. 20, No. 1, 77-97 (1992; Zbl 0764.62038)]. The estimators are shown to be consistent, and simulations are presented.