The principal investigator will investigate three aspects of statistical inference for time series models: iterative methods of estimation for incomplete dependent data, non-linear time series analysis, forecasting with a possibly misspecified model. The non-linear time series analysis represents a continuation of the investigator's work on a flexible class of threshold autoregressive models, where the relationship between stability and ergodicity of some time series models will be investigated. The work on estimation for incomplete dependent data will involve algorithm development. The principal investigator is working on various problems in time series analysis. Time series analysis is used to model situations were data are sequential and dependent. Current interest is to develope a class of models that account for non- linearities in real data. Another area of investigation will involve situations were data are incomplete. This again represents a situation that can occur in real data.
