The analysis of screening designs depends on the principle of effect sparsity, where most of the variation in the response is explained by a small number of effects. Under this principle, effects with small estimates are used to estimate the error in the model. This then allows one to test whether the larger effects are active.

Help created on 7/12/2018