Abstract
Some alternative tests to that of association are presented that are applicable when the data consist of the proportions of experimental units having a given attribute. In deriving the tests, treatment effects are assumed to be additive on the logit scale. Examples of tests for main effects and interactions are given by analyzing some data presented by Cochran. It is illustrated that once the P''s (the estimated proportions) are obtained under a certain no interaction hypothesis, the maximum likelihood estimates of the parameters in a linear model can be computed without iteration.
Keywords

This publication has 7 references indexed in Scilit: