Separable Statistics and Hypothesis Testing. The Case of Small Samples