ABSTRACT
If the attained standard treatment response is compatible with 50%, the null hypothesis, Ho:p ~ 0.50 is tested for each candidate treatment. Let X denote the number of responses observed among the N = 24 animals. A fixedsample-size design rejects Ho (ex = 0.05) if X is 17 or greater. The operating characteristics for this test are:
Type I error Power, p = 0.75 Power, p = 0.80
For logistical reasons the screening test is divided into three portions, with N = 8 animals tested within each portion. It would thus be simple to modify the fixed-sample-size test procedure to a three-stage test, with N = 8 animals per stage. The conduct and properties of this three-stage design were studied. The acceptance and rejection boundaries and the operating characteristics are as follows:
N 8 8 8 ak 3 10 16 rk 9 13 17
The specification of at = 3, rt = 9 implies that Ho can be accepted following stage 1, but not rejected. The acceptance and rejection boundaries following stage 3 coincide with the fixed-sample-size procedure.
