Durbin-Watson test statistic is computed in the context of the least weighted squares regression. The asymptotic distribution under the null hypothesis is investigated.
It turns out that it is asymptotically valid to carry out the Durbin-Watson test in the classical way also for robust regression.