Catanova Method for Determining of Zero Partial Association Structures in Multidimensional Contigency Tables
Zero partial association models correspond to the conditional independence relation of a variable pair, given the rest of variables in the model. For determining the model that best fits the data from the contingency tables measured at a nominal level, this study has used the C (CATANOVA) test statistic (with response as one of its variables, and factor as another one of its variables) instead of the chi-square statistic, which is used as the test statistic in Wermuth’s backward elimination method with zero partial associations. Numerical analyses were performed on two samples, and the associations between these statistics were evaluated. Interpretations were provided for the obtained results.