Fisher's exact test checks whether two categorical variables are associated, like a chi-square test of independence, but it computes the p value exactly instead of relying on a large-sample approximation. Use it when the sample is small: the usual rule is when more than 20% of the expected cell counts fall below 5. Because it has no test statistic, the APA 7 write-up is shorter than a chi-square one.
What you need before you write a single word
- The counts and percentages in each cell of the table.
- The expected counts, to justify using Fisher's test instead of chi-square.
- The exact p value (two-sided unless you planned a one-sided test).
- An effect size: the odds ratio for a 2 × 2 table, or φ.
The APA 7 format template
Because [X] of the [4] expected cell counts were below 5, Fisher's exact test was used. [Outcome] was significantly more common in [group 1] ([n] of [N], XX%) than in [group 2] ([n] of [N], XX%), p = .XXX.
All 53 templates, ready to paste with the italics set: the Rescue Vault, $27 →
Formatting rules reviewers actually check:
- Report no χ² value and no df: Fisher's test does not produce them.
- State that the p value is two-sided (the default) or one-sided.
- Give the percentages, so the direction of the association is clear.
- No leading zero on p.
A worked example
Suppose 19 patients received a new drug or a placebo, and you recorded whether each reported a side effect.
| Group | Side effect | No side effect | Total |
|---|---|---|---|
| Drug | 7 | 2 | 9 |
| Placebo | 1 | 9 | 10 |
Two of the four expected counts (3.79 and 4.21) are below 5, so a chi-square test is not appropriate. Fisher's exact test gives p = .005 (two-sided). Written up in APA 7:
Because two of the four expected cell counts were below 5, Fisher's exact test was used. Side effects were significantly more common in the drug group (7 of 9, 78%) than in the placebo group (1 of 10, 10%), p = .005 (two-sided).
You can add the odds ratio for effect size. With counts this small its confidence interval is very wide, so report it if your field expects it and say that it is imprecise.
Mistakes reviewers catch
- Inventing a χ² value. Fisher's test has none; report only the exact p.
- Using chi-square anyway. With expected counts below 5, the chi-square approximation is unreliable. Show the expected counts that justified your choice.
- No percentages. "There was a significant association" does not say which group was more likely to have the outcome.
- One-sided p without saying so. The default is two-sided; say so if you used a one-sided test, and why.
Let KyroStat do the write-up for you
KyroStat checks the expected counts, switches from chi-square to Fisher's exact test when they are too small, and writes the sentence from your data.
Frequently asked questions
When do I use Fisher's exact test instead of chi-square? When more than 20% of the expected counts are below 5, or any is below 1. Otherwise use the chi-square test of independence.
Does Fisher's exact test work for tables larger than 2 × 2? Yes (the Fisher-Freeman-Halton extension), and the write-up is the same: name the test and report the exact p.
Which effect size goes with it? The odds ratio for a 2 × 2 table; see how to report an odds ratio. For larger tables, Cramér's V.