George Mason University
CSI/Statistics Colloquium Series
Seminar Announcement


Simulation and Estimation of Fractionally Integrated Time Series

Jeff Butler

U.S. Internal Revenue Service


ABSTRACT

Fractionally integrated random processes are introduced and examined as an alternative to short-memory ARIMA models resulting from integer differencing. It is seen that fractionally integrated time series exhibit persistence at long lags of the autocorrelation function that allows for improved forecasting. Truncated and exact methods for synthetic generation are provided, examples are drawn, and a procedure for estimating fractional differencing parameter d is given. All computational algorithms are presented using the SAS System.


Friday, February 12, 1999
George W. Johnson Center, Assembly Room H
Note room change.
Seminar at 10:45 a.m.
Refreshments at 10:30 a.m.


Information about the Statistics Colloquium Series, including directions, and current and past schedules, is available at www.science.gmu.edu/statseminars