Ponente
Descripción
Entropy as presented by Shannon (1948) provides a simple way to express the uncertainty of random variables. The statistic SU, calculated by comparison of entropies, formally proposed in 2019, describes the correlation between two categorical variables X and Y. Statistical and mathematical properties of SU have been analyzed in subsequent works (Ortiz et al, 2026), showing attractive characteristics for usability in medicine, biology and other life sciences.
The chi-square test of independence provides a p-value which is high when independence is unlikely. Guessing that SU correlation and p-value may be related follows as a consequence.
This work studies how the SU values and the p-values of the chi-square test are related, for contingency tables of various sizes. We ran successive simulations, one for each contingency table configuration, recording in each case the observed SU and the observed p-value, and plotting these points in search of a relationship.
Results indicate, in most cases, a monotonic relationship between SU and p-value, giving further confirmation that SU is an adequate measure of correlation or dependence. A follow-up of this work will involve establishing similar advantages for MSU, the multivariate extension of SU.