Title:  Functional Coefficient Autoregressive Modeling of Multivariate Temporal Data

Presenter:  Dr. Jane L. Harvill, Mississippi State University

Univariate nonlinear time series models have been used extensively over the past 15 to 20 years to model complex dynamic systems that cannot be adequately represented using linear models.  A very general type of nonlinear time series model is the univariate functional coefficient autoregressive (FCAR) model, where the autoregressive coefficients are allowed to change as a function of one or more variables, which may be lagged values of the series or exogeneous predictors, including, for example, time.  We extend the univariate FCAR model to the vector time series framework.  A bootstrap test for vector time series nonlinearity is presented.  FCAR methods are applied to a number of different datasets to illustrate utility.  Extensions to spatial-temporal modeling are mentioned, but remain an open problem.