Description:
<p>We compare Geary's test based on the number of sign changes in residuals with the Durbin-Watson (DW) test for serial correlation. We derive the asymptotic distribution of Geary's statistic for the special case of identically distributed random variables and show that it is asymptotically distribution free for symmetric errors. Monte Carlo simulations show that the DW procedure is generally more powerful and robust than Geary's test.</p>