A Bootstrap Causality Test for Covariance Stationary Processes
|Speaker:||Javier Hidalgo, London School of Economics|
|Date:||Friday 25 October 2002|
|Location:||Room 106 Streatam Court|
This paper examines a nonparametric test for Granger-causality for a vector covariance stationary linear process under, possibly, the presence of long-range dependence. We show that the test converges to a non-distribution free multivariate Gaussian process. Since, in contrast to the scalar situation, it is not easy, if at all possible, to find a time transformation such that the transformed Gaussian process is a vector with independent Brownian motion components, inferences based on this process will be difficult to implement. To circumvent this problem, we propose to boot-strapping the tests by two alternative, although similar, approaches showing their validity and consistency.