Separation test

A separation test is a statistical approach for use in early-phase studies.

Separation tests are based on the Neyman-Pearson approach to hypothesis testing in statistics. They are distinct from, and have a different purpose than conventional null hypothesis tests. Whereas the usual tests provide a "confirm or reject" decision with respect to a hypothesis, separation tests result in a "continue-research or abandon-research or can't tell" result.

A simple separation test pertains to the estimate ${\hat {\theta }}$ of a parameter $\theta$ in a statistical model. Usually the parameter is defined so that $\theta$ is zero for a situation in which there is an absence of some effect, and values greater than 0 stand for a beneficial effect, and values less than 0 stand for a deleterious effect.

Further reading