Cramer's V

Definition

Cramer's V is a statistical measure used to show the degree of association between two categorical features. It is specifically tailored for categorical data, and is particularly useful when dealing with features that have multiple categories and you want to understand the strength of their association.

Range of score: 0 to 1

A score of 0 indicates no association between the features, suggesting that the categories of one feature do not provide any predictive information about the categories of the other feature. Conversely, a score of 1 signifies a perfect association, implying that knowing the category of one feature allows for perfect prediction of the category of the other feature.

How it works

Unlike many other statistical measures, Cramer's V is not influenced by the marginal distribution of the features. This means that it measures the association between the categorical features independently of the number of observations in each category. At its core it uses chi-square statistic that gets normalised by the sample size and minimum possible dimension of the contingency table.