When comparing two independent samples from the negative binomial distribution the test of equality of their expected values might be of interest. In this case, the assumptions on the nuisance shape parameter determine the form of suitable test statistic. This paper focuses on test based on Wald statistic under an assumption that the shape parameters in the two samples are known and equal. However, the common shape parameter might be replaced by its estimate, either statistical or expert. We discuss the consequences of this procedure on the power as well as on the size of the test (probability of rejection of H0 when it is true).