On threshold autoregressive processes
Jiří Anděl, Ivan Netuka, Karel Zvára · Czech digital mathematics library · 1984
A threshold autoregressive process of the first order with Gaussian innovations is analyzed in the paper.Four methods are used for finding its stationary distribution: (a) the exact solution is derived in some special cases, (b) a Markov approximation is proposed, (c) a numerical method for solving an integral equation is applied, (d) simulations are produced.In the cases (b) and (d) also estimates of the correlation function and of the spectral density are compared.