To analyze multi-cell tables (categorized data) we proposed a two-stage procedure characterized in application of Simonov-Tsai diagnostic at the first stage to evaluate the opportunity of subsequent analysis based on Pearson chi-square test. Comparing the obtained results of verification of statistical hypothesis of absent dependence between liner factor and column factor in a multi-cell table we found that criteria used for its verification: Pearson chi-square test, plausibility relation, powerdivergence family Cressie-Read, blended weight Hellinger and Zelterman’s statistic give close results. For calculation we used the data of DP questioning obtained from implementation of Harvard Project in 1948-1951. It was stated that the degree of disapproval of Soviet regime did not depend on their ethnic origin, qualification level, belonging to a certain social group, individual life circumstances in Soviet period.
categorized data, multi-cell tables, verification of statistical hypotheses, Simonov-Tsai diagnostic, Pearson chi-square test, plausibility relation criterion, power-divergence family Cressie-Read, blended weight Hellinger and Zelterman’s statistic
"Statystycheskyi analyz prychyn antysovetskykh nastroenyi u peremeshchёnnыkh lyts po dannыm Harvardskoho proekta" ,
Information Processing Systems,