A numerical comparison of two criteria for determining the order of ar processes
Authors:
Ottar Sandvin a;
Dag Tj
stheim b
stheim b
| Affiliations: | a 2007 Kjeller, NTNF/NORSAR, Norway |
| b Norwegian School of Business, Bergen, Norway |
DOI:
10.1080/02331888108801569
Publication Frequency:
6 issues per year
Subjects:
Mathematical Statistics;
Statistical Theory & Methods;
Statistics;
Statistics for the Biological Sciences;
Stochastic Models & Processes;
Formats available:
PDF
(English)
View Article:
View Article (PDF)
Abstract
We study two criteria for deciding the order of AR processes. One of the criteria is AKAIKE'S FPE criterion. The other criterion was introduced by G. SCHWARZ in a non-time series context. We translate SCHWARZ'S criterion to a time series situation and via simulation experiments we make a numerical comparison between the new criterion and the FPE criterion. The new criterion was found to outperform the FPE criterion for low order models, while the opposite seems to be true for higher order models.
|
| Keywords: Autoregrssive serics; order determination; FPE criterion; BTC criterion |
| view references (19) |

Download Citation


CiteULike
Del.icio.us
BibSonomy
Connotea