A Model for High-Order Markov Chains

Abstract

SUMMARY A model for Markov chains of order higher than one is introduced which involves only one additional parameter for each extra lag. Asymptotic properties and the autocorrelation structure are investigated. Three examples are given in which the model appears to model data more successfully than both the usual high-order Markov chain and the alternative models of Jacobs and Lewis (1978), Pegram (1980) and Logan (1981).