We consider non-numerical data to be organized in a form of a fourfold table. Such form is usual for two categorized factors, and every factor has two categories. In particular, such form of statistical data is widely used in medical, genetical, or psychological problems. Experts of those non-mathematical fields of science evaluate the factors colligation with the so-called relative risk coefficient. However, mathematical statistics operates with Pearson’s correlation coefficient or chi-squared statistics. In the paper, we investigate relations between theoretical and practical approaches to non-numeric factors colligation evaluation. It is shown that these methods in the most common cases provide the same (or similar) results, especially, in situations when the presumed colligations do not exist. For some uncommon cases (for example, when the table contains zeroes) we demonstrate the source of the differences and discuss advantages of the theoretical approach in such cases.