A bias-correction for Cramér’s and Tschuprow’s

Cramér's V and Tschuprow's T are closely related nominal variable association measures, which are usually estimated by their empirical values. Although these estimators are consistent, they can have large bias for finite samples, making interpretation difficult. We propose a new and simple bias correction and show via simulations that, for larger than 2 × 2 tables, the newly obtained estimators outperform the classical (empirical) ones. For 2 × 2 tables performance is comparable. The larger the table and the smaller the sample size, the greater the superiority of the new estimators.