Hewett J E, Bair E
Biometrics. 1986 Sep;42(3):647-51.
It is shown how two-stage methods developed for continuous distributions can be used to construct two-stage tests for certain hypotheses pertaining to discrete distributions. Specifically, a two-stage test for ordered means in the Poisson case is presented. An example from mutagenicity testing is discussed and data from it are used to illustrate the methodology. Results of a Monte Carlo power study are presented as well as a brief table of critical values.
