A Robust Test for Short Memory
|Speaker:||Brendan McCabe, University of Liverpool|
|Date:||Friday 14 February 2003|
|Location:||Room 106 Streatam Court|
(with S.J. Leybourne and D. Harris).
We suggest a non-parametric statistic to test the null hypothesis of short memory against the alternative of long memory. The statistic is simple to construct, being based on a single sample autocovariance whose order is a function of the sample size. We show that it has a standard normal limiting distribution under the null hypothesis, even when it is calculated from residuals from a fitted regression model with deterministic and/or stochastic regressors. We also demonstrate consistency of the test under long-memory alternatives. Monte Carlo simulations of empirical size and power suggest that the test will work well in practice.