Statistics Glossary → Odds ratio (OR)
Odds ratio (OR)concept
The ratio of the odds of an event in one group (the probability it happens divided by the probability it does not) to its odds in another group.
An event with probability p has odds p/(1 - p); if p = 0.20 the odds are 0.25 (1 to 4). For a 2x2 table with cells a, b, c, d, OR = (a·d)/(b·c). OR = 1 means no association, OR > 1 that the event is more likely in the first group, and OR < 1 that it is less likely. The confidence interval is usually built on the log scale (Woolf's method: SE = √(1/a + 1/b + 1/c + 1/d)); when a cell is zero, 0.5 is added to every cell (the Haldane-Anscombe correction). In logistic regression, the exponentiated coefficient exp(B) is the OR for that predictor with the other variables held constant.
The OR is often read as if it were a relative risk (RR); but when the event is common (above roughly 10%), the OR moves noticeably away from the RR and exaggerates the effect. Its advantages are that it can be estimated in case-control designs and that it is symmetric with respect to which category is called the 'event'. For paired data (McNemar), the conditional OR = b/c is computed from the discordant pairs only; for stratified data, a common OR is combined with the Mantel-Haenszel method.
An everyday example
After a meal at a camp, 20 of the 40 people who ate chicken fall ill, compared with 6 of the 60 who did not. The odds are 20/20 = 1 among chicken eaters and 6/54 ≈ 0.11 among the others, giving OR = 9. Yet the risks are 50% and 10%, so the relative risk is 5; because the event is common, the OR looks much larger than the RR.
In MerQur
A core output of the chi-square independence test, Fisher's exact test, crosstabs, Cochran-Mantel-Haenszel, the effect size calculator, meta-analysis, and logistic, multinomial, ordinal and conditional logit regression. It appears as 'Conditional OR CI' in the McNemar test, 'Cell association measures' in the MR × MR crosstab and 'Odds ratio / risk ratio for 2x2 tables' in weighted frequency; GEE gives an OR table with the logit link, and Confusion Matrix Metrics gives the diagnostic odds ratio (DOR).
Analyses where this concept appears
Chi-Square IndependenceFisher's Exact TestMcNemar TestCochran-Mantel-Haenszel (CMH) TestCrosstabMR × MR CrosstabEffect Size CalculatorMeta-AnalysisLogistic RegressionMultinomial Logistic RegressionOrdinal Logistic RegressionGeneralized Estimating Equations (GEE)Confusion Matrix MetricsWeighted Frequency (Survey Freq)Weighted Logistic Regression (Survey Logistic Regression)Conditional Logit (Choice Model)
Related concepts
Relative risk (RR) and risk difference (RD)Mantel-Haenszel common odds ratioHaldane-Anscombe (zero-cell) correction