Durbin-Watson Statistic


Description Method

Description

The Durbin-Watson statistic is used to perform the Durbin-Watson test for the absence of autocorrelation.  The software generates a Durbin-Watson statistucs for first-order autoregression.  The resulting value can be used in conjunction with a table of values to determine if the error terms are first-order autoregressive.  The resulting statistic can be tested against two values, dL and dH, based on the number of regressors and the number of degrees of freedom in the regression.  The null hypothesis, that there is no autoregression present, is tested as follows:

d < dL Reject null hypothesis
d > dH Do not reject null hypothesis
dL = dH Test is inconclusive

Method

The Durbin-Watson statistic is computed from the following equation:

d = sum((e(t) - e(t-1))^2)/sum(e(t)^2)
(from Kamenta, Jan, Elements of Econometrics: Second Edition, The University of Michigan Press, 2004, pp. 328-329.)
Copyright © 2007 Approximatrix, LLC
Text licensed under the Creative Commons Attribution-Share Alike 3.0 License
DracoTM and the Approximatrix logo are trademarks of Approximatrix, LLC
Other trademarks are property of their respective owners