Critical issues with the Pearson's chi-square test
Abstract
Pearson's chi-square tests are among the most commonly applied statistical tools across a wide range of scientific disciplines, including medicine, engineering, biology, sociology, marketing and business. However, its usage in some areas is not correct. For example, the chi-square test for homogeneity of proportions (that is, comparing proportions across groups in a contingency table) is frequently used to verify if the rows of a given nonnegative (contingency) matrix are proportional. The null-hypothesis : `` rows are proportional'' (for the whole population) is rejected with confidence level if and only if , where the first term is given by Pearson's formula, while the second one depends only on , and , but not on the entries of . It is immediate to notice that the Pearson's formula is not invariant. More precisely, whenever we multiply all entries of by a constant , the value is multiplied by , too, . Thus, if all rows of are exactly proportional then for any and any . Otherwise, becomes arbitrary large or small, as positive is increasing or decreasing. Hence, at any fixed significance level , the null hypothesis will be rejected with confidence , when is sufficiently large and not rejected when is sufficiently small, Yet, obviously, the rows of should be proportional or not for all simultaneously. Thus, any reasonable formula for the test statistic must be invariant, that is, take the same value for matrices for all real positive . KEY WORDS: Pearson chi-square test, difference between two proportions, goodness of fit, contingency tables.
Cite
@article{arxiv.2505.06318,
title = {Critical issues with the Pearson's chi-square test},
author = {Vladimir Gurvich and Mariya Naumova},
journal= {arXiv preprint arXiv:2505.06318},
year = {2025}
}