Testing for no autocorrelation using a modified Lobato test

Abstract

This paper suggests modifying the Lobato test for no autocorrelation by using the bandwidth parameter (M) of the covariance estimator as a fixed proportion of the sample size (T): M=bT, where b (0,1] is a constant. It is shown by means of simulations that the modified test has good control over size regardless the choice of b and a higher testing power can be achieved if a mall b is chosen.