Statistical model in tests for eye irritants.
A test used to classify substances for eye irritancy, as required by the Consumer Product Safety Commission, is performed on 1-3 groups of 6 albino rabbits in a sequential manner. When the statistical implications of the test are realized, it is possible for a substance to be classified as an irritant with fewer reactions than the number required for it to be classified as not an irritant. A procedure is given for correcting the inconsistency in the current test, and an alternative test, which considerably reduces the number of animals required, is proposed. Probability models and expected sample size calculations have been derived.