Volume 19, Issue 5 (September 1991)
Models and Interactions
It is shown that the usual analysis-of-variance model with interaction terms can be a hindrance to the understanding of data rather than a useful tool. An alternative model, especially useful for the analysis of interlaboratory test data, is discussed. However, the best treatment of certain types of data is obtained by an analysis that involves minimal prior assumptions. An example is discussed that illustrates the thesis of this paper.