Determination of model fitting with power-divergence-type measure of departure from symmetry for sparse and non-sparse square contingency tables

Abstract

In this study, we propose a gradation to decide the model fitting under the symmetry model by using the departure measure. An extensive simulation study under different parameter settings and scenarios is conducted to introduce a gradation on unit interval. The proposed gradation provides an opportunity to decide the model fitting under the symmetry model when the chi-squared assumption does not hold for small sample sizes. The usefulness of proposed gradation is proved empirically using four applications to real data sets.