Анотація:
The Neyman−Pearson’s problem of developing composit one sample mean test is considered. It is shown that this problem is reduced to the linear programming. The resulting test is general enough, since it does not require knowledge of underlying distribution, nor independence of observations.