A New Test for Autocorrelated Errors in the Linear Regression Model

Abstract

A test for autocorrelated errors in the linear model is introduced and shown to have, in general, greater power than the Durbin and Watson test for high values of autocorrelation.

Two Stage and Related Estimators and Their Applications

“…Apart from a two other choices are featured in the literature when regressors are exogenous. The first derives from Berenblut and Webb (1973). Consider first differencing (48) to produce Yt = x.a + u t where Yt = Yt -Yt-l and x t = x, -X t-l. A regression of Yt against x. provides a consistent estimator of a.…”

An empirical analysis of the UK Treasury bill market

“…the hypothesis 0 = 1 cannot be tested by conventional methods. R is the Berenblut and Webb (1973) statistic, shown by Sargan and Bhargava (1983a) to be the most powerful invariant test of 8 = 1, and it clearly rejects this 1 hypothesis.The L(2)statistic is the likelihood ratio test ofthe hypothesis/?, = /?, = 0,and this is I certainly not rejected. Furthermore, we note that, following Hendry and Richard (1982) the I random walk model variance dominates the Fama-Mundell-Tobin model.…”

Modification of the point optimal unit root test

“…On the other hand, under a unit root, Dufour and King (1991) recommend treating as a nuisance parameter. However, they end up employing the Berenblut and Webb (1973) assumption and set ¼ 1, which is also employed in LM unit root testing and by ERS and Elliott (1999). This assumption is also employed in this article.…”

“…2 Alternative assumptions (not examined in this article) include: u 1 ¼ " 1 random with E(" 1 ) ¼ 0 and finite variance and theBerenblut and Webb (1973) assumption u 1 ¼ " 1 $ N(0, 2 ).…”