|
Activity Number:
|
514
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Thursday, August 7, 2008 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Statistical Computing
|
| Abstract - #300680 |
|
Title:
|
On Sampling and Spurious Granger Causality
|
|
Author(s):
|
Victor Solo*+
|
|
Companies:
|
University of New South Wales
|
|
Address:
|
Kensington, Sydney, International, 2073, Australia
|
|
Keywords:
|
time series ; granger causality
|
|
Abstract:
|
We discuss the impact of sampling and additive noise on Granger causality. It is well known that in general a subsampled vector time-series (VTS) model is a vector ARMA (VARMA) process; but there has been no general procedure for computing it. Using state space methods based on the algebraic Ricati equation we provide such an algorithm. We then give examples to show that spurious Granger causality (GC) can arise due to slow sampling. We also show spurious GC can arise due to additive noise. The consequences of these results are profound. They suggest that without prior knowledge of the time scale of causality the results of any GC analysis must be in doubt. It is also implicit that GC analysis can only be done with VARMA models not with VAR models as is commonly the case.
|
