Time Series ARMA Model Identification by Estimating Information.
TEXAS A AND M UNIV COLLEGE STATION INST OF STATISTICS
Pagination or Media Count:
Statisticians, economists, and system engineers are becoming aware that to identify models for time series and dynamic systems, information theoretic ideas can play a valuable and unifying role. Models for time series Yt can be formulated as hypotheses concerning the information about Yt given various bases involving past, current, and future values of Y. and related time series X.. To determine sets of variables that are sufficient to forecast Yt, and especially to determine an ARMA model for Yt, an approach is presented which estimates and compares various information increments. The author discusses how to non-parametrically estimate the MAinfinity representation, and use it to form estimators of the many information numbers that might compare to identify an ARMA model for a univariate time series. Author
- Statistics and Probability