Abstract:
The paper deals with computation of critical values of Wald's test for comparison of means of two normal samples. Dependence of a significance level on nuisance parameters is studied for arbitrary sample sizes. The results due to Wald are generalized with the help of a Wald's test property based on the idea of interpolation. Critical values of the test and maximal and minimal values of the true significance level are tabulated as functions of sample sizes for nominal significance level $0.05$.