Abstract
Thie study develops some corrections of the normal distribution approximation test for difference beteween two Poisson parameters. This artivle deals with a comparison of type I error for differences among five tests (exact test, normal approximation test, Yates's correction for continuity, deliberate on kurtosis and square root-transformation), give the hypothesis that H_0 : λ_1=λ_2. The normal approximation test by Yate's correction for continuity provides a vert good approximation to the exact test characterized by Binomial distribution of P=1/2.
