Abstract
This article develops some corrections of the normal approximation test for use with the Poisson variable. The proposed corrections perform much better than the customary "Yates' correction." This paper performs a function similar to type I error as a means of estimating differences between two tests ( the Poisson test and the normal approximation test), given the hypothesis that λ=λ_0. The true type I errors are much nearer to the nominal values for any value of λ using the corrections proposed here than using the "Yates' correction."
