|
Activity Number:
|
361
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Wednesday, August 6, 2008 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Nonparametric Statistics
|
| Abstract - #301991 |
|
Title:
|
Consistent Nonparametric Tests for Granger Causality
|
|
Author(s):
|
Yoshihiko Nishiyama and Kohtaro Hitomi and Yoshinori Kawasaki*+ and Kiho Jeong
|
|
Companies:
|
Kyoto University, Institute of Economic Research and Kyoto Institute of Technology and The Institute of Statistical Mathematics and Kyungpook National University
|
|
Address:
|
4-6-7 Minami-Azabu, Minato-ku, Tokyo, International, 106-8569, Japan
|
|
Keywords:
|
time series ; nonlinear causality ; nonparametric test ; local alternatives
|
|
Abstract:
|
Most of the existing tests of Granger-type causality are concerned only with "linear causality in mean", namely when a series linearly affects the (conditional) mean of the other series. However, dependence between series may be nonlinear, and/or not only through the conditional mean. The purpose of this paper is to propose nonparametric tests for possible nonlinear causality up to K-th conditional moment. A desirable property of the tests is that they have nontrivial power against square-root-T-local alternatives, where T is sample size. Their null asymptotic distributions are not normal, but we can easily calculate the critical regions by simulation. Monte Carlo experiments show that the proposed tests have good power properties, and much better than the competitors. An application is given in the context of causality in conditional heteroscedasticity.
|
