Estimation in Nonlinear Time Series Model II: Some Nonstationary Series

摘要

In an earlier paper a general framework was introduced for analyzing estimates in stationary nonlinear time series models. In this present paper the framework is enlarged to include certain nonstationary and nonlinear series. General conditions for strong consistency and asymptotic normality are derived both for conditional least squares and maximum likelihood type estimates. Examples are taken from threshold autoregressive, random coefficient autoregressive and doubly stochastic (dynamic state space) models. The emphasis in the examples is on conditional least squares estimates. (Author)