'A test for normality based on the empirical characteristic function'

Abstract

The last four rows of Table 1 in this paper are in error, due to a mis-set parameter in the simulations. New, extensive simulations show that the empiric characteristic function method, column A, tends to be superior against contaminated normal CN(α,σ2) alternatives when σ2 is small, but tends to be inferior when σ2 is large. We give below corrected versions of the 4th last and 3rd last rows of Table 1, and two new rows which indicate behaviour for smaller values of σ2. The notation follows that of the original paper. Show more