An empirical study of the power of the Terpstra-Jonckheere test using binary data.

A Monte-Carlo simulation is performed to evaluate the power of the Terpstra-Jonckheere test for monotonic trend as applied to binary data. The effects of the magnitude and shape of the underlying response and of varying the number of treatment levels for in vitro experiments in cytogenetics are examined. It is shown that choice of the number and spacing of the treatment levels can be critical factors in determining the power of the test. These findings indicate that selection of the proper experimental design is a very important step in the process of conducting a viable study.