SPSS CROSSTABS statistics: chi-square, Fisher, kappa and McNemar

In Crosstabs, build the table in Table mode and open All statistics to get the full set: chi-square tests, Fisher's exact tests, nominal and ordinal measures with asymptotic standard errors, correlations, kappa and McNemar, risk estimates, Cochran–Mantel–Haenszel tests across a layer, and cell residuals.

• Whether the row and column variables are independent: Pearson chi-square, continuity correction, likelihood ratio, and Fisher's exact test for 2×2 and larger tables.
• How strong the association is for nominal variables: phi, Cramér's V, contingency coefficient, lambda, Goodman–Kruskal tau and the uncertainty coefficient.
• For ordered categories, the direction and strength of the trend: gamma, Kendall's tau-b and tau-c, Somers' d, Spearman and Pearson correlations and the linear-by-linear test.
• Agreement between paired ratings (Cohen's kappa, McNemar or McNemar–Bowker) and, in 2×2 tables, odds ratios and relative risks with confidence intervals.
• Whether a 2×2 association holds across the levels of a third variable: Cochran's and Mantel–Haenszel tests, Breslow–Day and Tarone homogeneity tests and the common odds ratio.

1. 01

   Open the data in Table mode

   Load your file in the workspace and open Table mode, where tables and their statistics are built.

2. 02

   Assign the row and column variables

   In the variable list, click Row on one variable and Col on another. The table and its chi-square summary appear below it.

3. 03

   Declare measurement levels

   Ordinal measures and correlations need both variables ordered: numeric codes count as ordered, or declare text categories ordinal in the data dictionary. Kappa and McNemar appear for square tables whose rows and columns share categories.

4. 04

   Open All statistics

   Expand All statistics (G-test, Fisher's exact, λ, γ, τ, d, U…) under the table to see the chi-square tests, nominal and ordinal measures, correlations, agreement and risk estimates.

5. 05

   Add a layer and residuals

   In the Layer analysis section, choose a Layer (control) variable for the Cochran–Mantel–Haenszel tests. Under Display, set Residual in cells to Raw, Std. or Adjusted to show residuals in the table.

Chi-square tests

Pearson chi-square, continuity correction (2×2), likelihood ratio, Fisher's exact test (one- and two-sided for 2×2), the Fisher–Freeman–Halton exact test for larger tables with small expected counts, and linear-by-linear association for ordered variables.

Nominal measures

Phi, Cramér's V and the contingency coefficient with p, then lambda, Goodman–Kruskal tau and the uncertainty coefficient (symmetric and each direction) with asymptotic SE, approximate T and p.

Ordinal measures

Gamma, Kendall's tau-b and tau-c, and Somers' d (symmetric and each direction) with asymptotic SE, approximate T and p; withheld with the reason when a variable is not ordered.

Correlations

Pearson's R on category scores and the Spearman correlation on midranks, plus eta when a variable is interval-level.

Agreement

Cohen's kappa with SE and p, and the McNemar (2×2, exact binomial) or McNemar–Bowker symmetry test.

Risk estimate

For 2×2 tables: the odds ratio and the relative risk for each column category, with 95% confidence intervals.

Layer analysis

With a layer variable and a 2×2 table in each layer: Cochran's and Mantel–Haenszel tests, Breslow–Day and Tarone tests, and the Mantel–Haenszel common odds ratio with each layer's odds ratio.