An investigation of the method of accumulation analysis

‘Accumulation analysis’ is a method introduced by Taguchi (1974) for the analysis of ordered categorical data. Using his original data on the efficacy of two rival drugs, we discuss the reasons Taguchi gives for substituting his accumulation statistic for Karl Pearson's χ2statistic. We show that the accumulation statistic gives greater weight to the linear component of χ2. We point out the superiority and greater simplicity of a well-known procedure employing this linear componentto determine the efficacy of a treatment. Finally, using a result due to Yates, we show that the extraction of the linear component of χ2is itself equivalent to simply analysing equally-spaced scores and conducting a standard analysis using the statistic or the analysis of variance. In view of these results and the findings of earlier research by ourselves, Nair (1986), and Hamada & Wu (1986), we conclude that accumulation analysis is inefficient, as well as unnecessarily complicated, and consequently should not be taught or reco...