Statistics Glossary → Ordinal association measures (gamma, tau-b, tau-c, Somers' d)
Ordinal association measures (gamma, tau-b, tau-c, Somers' d)concept
Coefficients that measure the direction and strength of the association between two ordinal categorical variables from the counts of concordant and discordant pairs of observations.
A pair is concordant (C) if one observation is higher on both variables, and discordant (D) if it is higher on one and lower on the other. All these measures scale the difference C - D in different ways and range from -1 to +1; a positive value means that as one variable increases, the other tends to increase too. Goodman and Kruskal's gamma = (C - D) / (C + D) ignores tied pairs altogether and therefore usually gives the largest value.
Kendall's tau-b accounts for ties in the denominator and suits square tables; tau-c (Stuart's) is adjusted so that it can reach ±1 in rectangular tables. Somers' d is asymmetric: it treats one variable as dependent and adds to the denominator only pairs tied on the dependent variable, so it gives two different values depending on direction. These measures capture only monotonic relationships; for an inverted-U relationship they can be close to zero.
An everyday example
A survey asks about education level (primary, middle, high school, university) and how often people read a newspaper (never, sometimes, often). If gamma = 0.45, then for two randomly chosen people who are not tied on either variable, it is clearly more likely that the more educated one also reads more often than the reverse.
In MerQur
The 'Ordinal measures' checkbox of the Chi-Square Independence test (gamma, tau-b, tau-c, Somers' d; CI from ASE1, z and p from ASE0); 'Additional outputs: Association measures' in the Crosstab.
Analyses where this concept appears
Chi-Square IndependenceCrosstab
Related concepts
Kendall's tauProportional reduction in error measures (lambda, Goodman-Kruskal τ, uncertainty coefficient)