Русский
!

Conference publications

Abstracts

XIX conference

Fractional integration models for time series with long memory

Grebenuk E.A.

V.A. Trapeznikov Institute of Control Sciences Russian Academy of Sciences, Profsouznaya 65, Moscow, 117997, Russia

1 pp. (accepted)

The standard unit root (difference –stationary –DS) and trend stationary (TS) framework has not capable to describe all the financial and macroeconomic indexes which may happen in the market. If DS-models describe the processes with unlimited memory, the TS – models describe the processes with finite memory, then significant number processes with long-term memory ought not to describe with DS or TS models satisfactorily. Long-term memory processes, and in particular models based on fractional integration as [1, 2] have come to play an increasing role in time series as they have yielded an especially large number of applications. The aim of this work to consider and compare the different approaches to specification model of the financial and macroeconomic indexes: DS or TS model setting and fractional integration setting. There are two major aspects of the time series analysis: testing of cointegration and detection of the dynamic changes – structural shifts. There are some publication in which the fractional integration approach advantages have been showed. We gave attention on the second approach. The structural shifts need detect as soon as possible after change have been occurred. The algorithms for sequential change detect of the DS processes were proposed in [3]. The main Russia macroeconomic indexes analysis revealed that the most of this may be describe with fractional integration models. The developed algorithm for sequential detect of the fractional integration processes departures from its model estimate the structure shifts times with great precision and small delay. The comparative analysis results for Russian macroeconomic index dynamics based on different models and corresponding its sequential change detect algorithms have been produced.

References

1. Granger C. W. J., Joyeux R. An Introduction to Long -Memory Time Series Models and Fractional Differencing// Journal of Time Series Analysis No. 1,1980. Pp. 15–29.

2. Hoskins J. R. M. Fractional Differencing // Biometrica No. 68, 1981. Pp. 165–176.

3. Grebenuk Е.А. Monitoring of stationary and non-stationary processes: on - line detection of the structural shifts//The second International Workshop in Sequential Methodologies (IWSM) Troyes, France, June 15-17, 2009, IWSM67.



© 2004 Designed by Lyceum of Informational Technologies №1533