Diagnostic Tests for Similar Action of Two Compounds

This paper proposes a class of diagnostic tests for the hypothesis of simple similar action of two drugs. The likelihood score test is shown to have a useful interpretation as the usual test for adding a variable to an appropriate weighted linear regression. This suggests use of the corresponding added variable plot as an auxiliary diagnostic. A detailed exposition is given for the situation of a continuous response with normal errors and for the quantal response case assuming binomial variation. The proposed methods are illustrated using data involving mixtures of two forms of insulin and data from an experiment to study the joint administration of two insecticides.