2004 年 25 巻 1 号 p. 1-18
This paper proposes to use the notion, composite maximum contrast method, for formulating recently proposed decision procedures for judging toxicity of chemical substances based on in vitro experiments. It is defined as the composite procedure which judges the substance in question to be positive if and only if two or more maximum contrast methods, such as Dunnett test and trend test, simultaneously realized statistical significance with a certain significance levels. SAS/IML programs are also introduced for calculating type I and/or type II error probabilities of composite maximum contrast methods. A real example of data analysis in a validation study of in vitro BALB/c 3T3 cell transformation assay, to which a composite maximum contrast method is applied, is also shown.