MerQuR+Statistics Glossary

Statistics Glossary → Cramér's V and Cohen's w

Cramér's V and Cohen's wconcept

Effect sizes derived from the chi-square statistic that show the strength of the association between two categorical variables (or of the gap between observed and expected distributions).

Because the chi-square statistic grows with sample size, it does not indicate the strength of an association. Cohen's w = √(χ² / n) removes the sample size from chi-square. Cramér's V = √(χ² / (n (k - 1))) also scales by the size of the table, where k is the smaller of the numbers of rows and columns. V ranges from 0 (no association) to 1 (perfect association); in a 2×2 table it equals the absolute value of the phi coefficient.

Cohen's benchmarks of 0.1 / 0.3 / 0.5 (small / medium / large) for w cannot be carried over to V directly: the same V corresponds to a stronger association in larger tables, so the benchmarks for V should be divided by √(k - 1). V has no direction; it does not show the form of the association or which cells drive it, for which the standardized residuals are examined.

An everyday example

A survey of 1,000 people looks at the link between preferred transport (bus, metro, car) and area of residence (centre, outskirts). χ² = 40 and p is tiny, but V = √(40 / 1,000) = 0.20. The association is significant, yet its strength is only small to medium.

In MerQur

The 'Cramér's V CI + Cohen interpretation' checkbox of the Chi-Square Independence test; 'Effect: Cohen w and V' in Chi-Square Goodness of Fit. Also reported in the Crosstab and the Effect Size Calculator.

Analyses where this concept appears

Chi-Square IndependenceChi-Square Goodness of FitCrosstabEffect Size Calculator

Related concepts

Phi (φ) coefficientEffect size