Title: On separation of symmetry for ordinal categorical data
Authors: Kouji Tahata - Tokyo University of Science (Japan) [presenting]
Abstract: Let us consider some trials which have more than two possible outcomes with ordinal categories. Then, the counts have the multinomial distribution. A problem of modeling of asymmetry for multinomial parameters with respect to the midpoint of categories is treated. Also, it is proved that the symmetry of multinomial parameters can be separated into the asymmetric structure of multinomial parameters and the structure of moment. This result can be applied for the multi-way contingency tables. For example, the decomposition of the point-symmetry mode was given time ago. Using these results, we can obtain a more detail decomposition of point-symmetry.