Statistics Glossary → Phi (φ) coefficient
Phi (φ) coefficientconcept
A measure of association that expresses the strength of the relationship between two binary (two-category) variables on a scale from -1 to +1.
In a 2×2 table φ = √(χ² / n), where χ² is the Pearson chi-square statistic without continuity correction. In terms of the cell counts the signed version is φ = (ad - bc) / √((a+b)(c+d)(a+c)(b+d)), and when the two binary variables are coded 0-1 it equals the Pearson correlation coefficient. φ = 0 indicates no association and ±1 a perfect one; Cohen's general benchmarks of 0.1 small, 0.3 medium and 0.5 large are commonly used.
φ is meaningful only for 2×2 tables; in larger tables its value can exceed 1, so Cramér's V is used instead (for a 2×2 table, Cramér's V equals |φ|). When the marginal distributions are very unbalanced, the largest value φ can reach is below 1, so comparing φ values across tables with different margins calls for caution. Like the odds ratio, φ summarizes the direction and strength of the association, but on a correlation scale rather than a ratio scale.
An everyday example
A school asks 200 pupils whether they eat breakfast and whether they have trouble concentrating in the morning lesson. Most of those who eat breakfast have no trouble, while a sizeable share of those who skip it do. If chi-square comes out at 18, then φ = √(18 / 200) = 0.30, pointing to a moderate association between the two yes-or-no variables.
In MerQur
It is requested with the 'φ, 2x2' checkbox in Chi-Square Independence and with the 'φ Cohen interpretation' checkbox in Fisher's Exact Test. MR × MR Crosstab gives a signed phi for each cell, and the Effect Size Calculator computes φ and Cramér's V with a noncentral χ² based confidence interval.
Analyses where this concept appears
Chi-Square IndependenceFisher's Exact TestMR × MR CrosstabEffect Size Calculator