MerQuR+Statistics Glossary

Statistics Glossary → Contingency coefficient C

Contingency coefficient Cconcept

A measure derived from the chi-square statistic that expresses the strength of association between two categorical variables on a scale from 0 to an upper bound below 1: C = √(χ² / (χ² + n)).

Because the chi-square statistic grows with sample size, on its own it does not show the strength of an association. Pearson's contingency coefficient C scales chi-square by the sample size, giving a measure between 0 (independence) and values approaching 1. It can be applied to tables of any size and carries no information about direction.

An important limitation of C is that it cannot reach 1 even under perfect association; its maximum depends on the size of the table (√((k - 1) / k) for a k x k table, about 0.71 for a 2 x 2 table). C values from tables of different sizes therefore cannot be compared directly. For this reason Cramér's V, which is scaled fully between 0 and 1, is usually preferred; dividing C by its maximum possible value is one way to ease interpretation.

An everyday example

A survey asks 500 people both their education level (three categories) and their main news source (three categories), and finds χ² = 60. Then C = √(60 / 560) ≈ 0.33. Since C can be at most about 0.82 in a 3 x 3 table, this indicates a moderate association, roughly 40% of the strongest possible one.

In MerQur

It is requested with the 'C, contingency coefficient' checkbox in Chi-Square Independence as C = √(X²/(X² + n)); in Crosstab, the 'Additional outputs: Association measures' option gives the contingency coefficient C together with measures such as gamma, tau and lambda.

Analyses where this concept appears

Chi-Square IndependenceCrosstab

Related concepts

Phi (φ) coefficientCramér's V and Cohen's w