Bayesian Cross-Tabulation Analysis
Bayesian Contingency Table Analysis · Also known as: Bayesian chi-square test, Bayesian contingency table test, Bayes factor association test, Bayesian crosstab analysis
Bayesian cross-tabulation analysis tests whether two categorical variables are associated by computing a Bayes factor that quantifies the evidence for an association model against an independence model. Unlike classical chi-square testing, it provides a continuous measure of evidence, supports the null hypothesis directly, and updates naturally with prior knowledge about the cell probabilities.
Read the full method
Sign in with a free account to read this section.
Method map
The neighbourhood of related methods — select a node to explore.
When to use it
Use Bayesian cross-tabulation when you have two or more categorical variables arranged in a contingency table and want to quantify evidence for or against their association. It is especially valuable when you need to support the null hypothesis of independence (which classical tests cannot do), when sample sizes are small, or when you want to incorporate prior knowledge. Do not use it when your primary audience requires classical p-values for regulatory or editorial reasons without Bayesian alternatives. It is not appropriate for continuous or ordinal outcomes where rank-based or regression models are more suitable.
Strengths & limitations
- Provides evidence in favour of the null hypothesis (independence), not just against it.
- Bayes factor is a continuous, calibrated scale of evidence rather than a binary reject/fail decision.
- Analytically tractable under the Gunel–Dickey Dirichlet-multinomial model — no simulation needed for standard tables.
- Naturally accommodates sequential updating as new data arrive.
- Less sensitive to large sample sizes inflating trivially small effects into significant results.
- Choice of prior (concentration hyperparameter) can influence Bayes factors, especially with sparse tables.
- Interpretation of Bayes factor thresholds (e.g., Jeffreys' scale) is conventional, not absolute.
- Less familiar to reviewers and editors in many applied fields where classical p-values remain the standard.
- Computationally more involved for high-dimensional tables or non-standard sampling designs.
Frequently asked
How does Bayesian cross-tabulation differ from the classical chi-square test?
The chi-square test computes a p-value — the probability of data at least this extreme if independence were true — and rejects H0 if p < 0.05. Bayesian cross-tabulation computes a Bayes factor, which directly compares the probability of the data under the association model to its probability under the independence model. This allows you to quantify evidence for independence itself, not only against it.
What does a Bayes factor of 5 mean?
BF10 = 5 means the data are five times more probable under the association model than under the independence model. Following Jeffreys' conventional scale, values between 3 and 10 are moderate evidence for association. The reciprocal BF01 = 0.2 would indicate moderate evidence for independence.
Which prior should I use?
The default Gunel–Dickey prior with concentration parameter alpha = 1 is recommended for most applied analyses. It is uniform over the simplex, requires no subjective input, and is the default in JASP. If you have genuine prior knowledge about expected cell proportions, a sensitivity analysis across a range of alpha values is advisable.
Can I use this for tables larger than 2x2?
Yes. The Gunel–Dickey framework extends to r x c tables of any size. However, sparse cells (expected counts below 1) make both prior specification and interpretation more difficult; consider collapsing categories or using an exact method in those cases.
Is the Bayesian cross-tabulation available in standard software?
Yes. JASP provides a point-and-click interface with default Gunel–Dickey Bayes factors. The BayesFactor package in R offers the same analysis with greater flexibility for prior specification and output format.
Sources
- Gunel, E., & Dickey, J. (1974). Bayes factors for independence in contingency tables. Biometrika, 61(3), 545–557. DOI: 10.1093/biomet/61.3.545 ↗
- Jamil, T., Ly, A., Morey, R. D., Love, J., Marsman, M., & Wagenmakers, E.-J. (2017). Default Gunel and Dickey Bayes factors for contingency tables. Behavior Research Methods, 49(2), 638–652. DOI: 10.3758/s13428-016-0739-8 ↗
How to cite this page
ScholarGate. (2026, June 3). Bayesian Contingency Table Analysis. ScholarGate. https://scholargate.app/en/statistics/bayesian-cross-tabulation-analysis
Which method?
Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.
- Bayesian chi-square testStatistics↔ compare
- Bayesian Independent Samples t-testStatistics↔ compare
- Chi-square testStatistics↔ compare
- Cross-tabulation analysisStatistics↔ compare