Fishers Exact Test Calculator - 2x2 p-Value and Odds Ratio

A fishers exact test calculator is a statistical tool that returns an exact p-value, sample odds ratio, and corrected confidence interval for the association between two categorical variables in a 2x2 table, without relying on the chi-square approximation.

• • Clinical and biomedical studies: Test whether a new treatment changes the odds of an outcome when one arm of a randomized trial enrolls fewer than 30 participants.
• • Genetics and genomics: Compare allele or genotype counts between cases and controls when expected cell counts are below five.
• • Marketing and survey research: Evaluate whether a binary response (clicked, purchased, recommended) differs between two audience segments of unequal size.
• • Quality engineering and Six Sigma: Confirm whether a defect category is concentrated in one shift, machine, or supplier when the daily counts are low.

The procedure works by treating the row and column totals as fixed, then computing the exact hypergeometric probability of every possible rearrangement of the four cells.

That conditioning is what makes the test 'exact' even when samples are tiny, because the p-value is not an approximation to a chi-square or normal distribution.

Once you paste or type the four observed counts, the calculator reports the exact p-value, the observed table's hypergeometric probability, the cross-product odds ratio, and a Wald-style 95% confidence interval for the population odds ratio.

When the expected counts are comfortably above five, many readers reach for the chi-square calculator first to get an approximate answer in seconds.

The engine enumerates every 2x2 table that can be built from your row and column margins, computes each table's hypergeometric probability, and sums the probability mass that is at least as extreme as your observed table.

• a, b, c, d: Observed counts in the 2x2 cells (top-left, top-right, bottom-left, bottom-right).
• n: Grand total equal to a + b + c + d, which must be greater than zero.
• row and column margins: Fixed by the data; the calculator treats them as conditioning totals when summing probabilities.
• tails: Two for the standard two-sided inference, one if you pre-registered a directional hypothesis.

Because every possible table shares the same margins, the denominator of the hypergeometric formula stays constant and only the numerator changes, which keeps the calculation fast even when totals exceed 100.

For two-sided inference the calculator sums the probabilities of every table whose probability is no greater than the observed table, the convention recommended by the NIST/SEMATECH handbook and most biostatistics textbooks.

For one-sided inference the calculator sums probabilities only in the chosen direction (odds ratio less than 1 or greater than 1), matching the convention used in clinical trial pre-registration.

According to NIST/SEMATECH e-Handbook of Statistical Methods, Fisher's exact test evaluates the association between two categorical variables using the exact hypergeometric distribution of cell counts with fixed margins.

According to GraphPad QuickCalcs - Analyze a 2x2 contingency table, the summing-small-p-values method used for Fisher's exact test sums the hypergeometric probabilities of every admissible table whose probability is no greater than the observed table, and is the recommended choice for small sample sizes.

Because the test reduces to a sum of hypergeometric terms that mirror the binomial distribution calculator, students often rehearse the logic on a single-row setup before tackling the full 2x2 case.

Before you interpret a p-value it helps to be clear on four ideas that show up every time you read a Fisher's exact test result.

These four ideas also explain why the result depends only on the four cell counts and not on the marginal rates you might compute from a larger study.

The 0.5-corrected logit interval uses the same normal-approximation reasoning that the confidence interval calculator applies to a single proportion, which makes that page a useful complement when you want to see how the Wald-style interval behaves with very small n.

Enter your four counts, choose whether you pre-registered a direction, and read the p-value alongside the odds ratio and its interval.

1. 1 Enter the four cell counts: Type the non-negative integer counts for a, b, c, and d exactly as they appear in your 2x2 table, leaving the defaults in place for the worked example.
2. 2 Choose one- or two-sided inference: Keep two-sided unless you wrote a directional hypothesis in your study plan before looking at the data.
3. 3 Pick a direction for one-sided tests: Set the direction to greater or less to match the alternative you pre-registered. The calculator ignores this field when two-sided is selected.
4. 4 Read the p-value first: Compare the p-value to your alpha threshold (commonly 0.05). Report it to four decimals unless it is below 1e-4, in which case use scientific notation.
5. 5 Interpret the odds ratio and its interval: An interval that excludes 1.0 reinforces the p-value decision; an interval that crosses 1.0 means the data are compatible with no association at the chosen confidence level.
6. 6 Reset and reuse: Use the Reset button to restore the default 8, 2, 1, 5 example and re-run with new counts for the next contingency table.

A clinical team runs a pilot trial with 16 enrolled patients: 8 of 10 in the treatment arm reach the endpoint and 1 of 6 in the control arm does. Entering a = 8, b = 2, c = 1, d = 5 returns a two-sided p-value of 0.0349 and a sample odds ratio of 20.0, which they can quote with the 95% interval when reporting the pilot.

Pairing a categorical analysis like this with a continuous-outcome check using the t-test calculator is a common workflow when a pilot reports both binary and numeric endpoints.

Using a purpose-built fishers exact test calculator saves time and avoids the silent errors that creep in when the test is done by hand or with a generic chi-square tool.

• • Reliable at small sample sizes: Produces an exact p-value even when expected counts are below 5, where the chi-square approximation breaks down.
• • One- and two-sided inference: Supports both the standard two-sided convention and pre-registered one-sided hypotheses, so the test matches your study design rather than forcing a default.
• • Odds ratio with corrected confidence interval: Returns the cross-product odds ratio alongside a Wald-style 95% interval with a 0.5 continuity correction, so you can quote both the point estimate and its uncertainty in the same view.
• • Educational transparency: Shows the observed probability of the input table, which makes it easier to explain to students why a small table can still produce a significant p-value.
• • Time savings for analysts: Avoids re-implementing the hypergeometric enumeration in R, Python, or a spreadsheet for every new contingency table.

These advantages are most visible in classroom settings where a single instructor needs to grade twenty 2x2 tables per assignment, and in pilot studies where regulatory reviewers expect exact inference alongside any chi-square summary.

Diagnostic studies that report a significant p-value here often pipe the same 2x2 counts into the post-test probability calculator to translate the association into a clinically meaningful probability shift.

Several data characteristics and modeling choices influence how you should read a Fishers exact test result.

• • Limited to 2x2 tables; for r x c tables larger than 2x2, Freeman-Halton or Monte Carlo extensions of Fisher's exact test are required.
• • Assumes the margins are fixed by design. If your study sampled both margins freely, condition on the observed margins only as an approximation and consider an unconditional exact test.

Most classroom and pilot-study use cases meet these conditions; the limitations matter mainly when you extend the test to multi-way tables or to designs where the margins themselves carry information.

According to R documentation: fisher.test, for 2x2 tables p-values come from the (central or non-central) hypergeometric distribution with the odds ratio as the non-centrality parameter, the same fixed-margin conditioning the calculator uses for its exact p-value.

Counting admissible tables under fixed margins is the same conditioning used by the allele frequency calculator when a genetics study tests whether observed genotype counts differ from Hardy-Weinberg proportions, the original use case Fisher designed the exact test for.