Fishers Exact Test Calculator

Background.

Fishers Exact Test Calculator turns a clearly defined set of inputs into a reproducible result using the standard variant named on this page. The implementation fixes the row and column margins, computes each hypergeometric table probability, and uses the standard probability-ordering convention for a two-sided p-value: sum every feasible table no more likely than the observed table. The central design choice is explicit because a useful calculator must tell you which question it answers before it shows a number. Every displayed input belongs to that chosen model, every result is computed from those inputs, and arithmetic is retained at full Decimal precision until the final output boundary.

Use the calculator by entering measurements in the units printed beside each field and by selecting only options that describe the case being analysed. The default values form a worked example, not a recommended or typical case. Change them to your own values and check that every unit, category and time basis matches the source information you have. A result with the wrong units can look plausible while answering a different question, so unit agreement is part of the calculation rather than presentation polish.

The implemented relationship is P(table) = C(r₁,a)C(r₂,c)/C(n,c₁); two-sided p sums tables with P≤Pobserved. The calculator validates each field before evaluating that relationship. Empty text, non-numeric values, non-finite numbers, values outside the stated physical or scoring domain, impossible ordering and unsupported modes produce a field-specific error instead of a fabricated result. Where a category or threshold is involved, comparison is made against the unrounded value unless the governing source expressly defines a rounded comparison.

Scope is deliberately narrow: The page explicitly chooses the probability-ordering two-sided convention. Mid-p, doubled one-sided and alternative conditional orderings are not reported, and the sample odds ratio is not a confidence interval. That limit is shown here, before the result is trusted, because a caveat hidden in a frequently asked question arrives too late. A different convention may also be defensible, but it is a different calculation and should be named separately rather than blended into an ambiguous output. The page therefore favours one auditable standard variant over an unexplained menu of approximations.

The primary reference is R Core Team, R stats manual current 2026, fisher.test two-sided probability ordering. It was checked against NIST/SEMATECH rather than copied from another calculator. The independent check agrees with the implemented equation and unit direction. Source identity, revision and locator are recorded so a later reviewer can tell whether a changed standard, new model revision or different population requires an update.

Worked examples on this page are executable fixtures. Their inputs are passed through the same registered formula used by the live widget, and the narrative states the resulting primary output rather than relying on a remembered calculation. This protects against a common publishing error in which correct code and hand-written prose quietly disagree. Secondary outputs are breakdowns of the same calculation and should reconcile with the primary value.

Treat the result as an estimate whose precision cannot exceed the inputs. More displayed digits do not repair approximate measurements, estimated densities, subjective score components, model calibration limits or contract assumptions. Run sensible low and high cases when an input is uncertain, keep the source values with your record, and ask a qualified professional to review any decision with safety, clinical, legal, structural or substantial financial consequences.

Finally, this page does not claim universal coverage. It does one named calculation transparently, cites the authority used, states what it leaves out and fails visibly when the inputs fall outside the model. That combination is more useful than a broader page that silently chooses assumptions for the user.

What is fishers exact test calculator?

Fishers Exact Test Calculator is a focused implementation of P(table) = C(r₁,a)C(r₂,c)/C(n,c₁); two-sided p sums tables with P≤Pobserved. The implementation fixes the row and column margins, computes each hypergeometric table probability, and uses the standard probability-ordering convention for a two-sided p-value: sum every feasible table no more likely than the observed table. The page explicitly chooses the probability-ordering two-sided convention. Mid-p, doubled one-sided and alternative conditional orderings are not reported, and the sample odds ratio is not a confidence interval.

How to use this calculator.

1. 01Read the scope statement and confirm that the named variant matches your question.
2. 02Enter every value in the unit printed beside its field and choose the applicable mode.
3. 03Resolve any field error rather than forcing an out-of-domain value through the formula.
4. 04Read the primary result together with its supporting outputs and the limitation beside it.
5. 05Save the inputs, source revision and date when the result informs a consequential decision.

The formula.

The implementation fixes the row and column margins, computes each hypergeometric table probability, and uses the standard probability-ordering convention for a two-sided p-value: sum every feasible table no more likely than the observed table.

Rounding stage: arithmetic remains at full Decimal precision and numeric results are rounded only at the return boundary. The page explicitly chooses the probability-ordering two-sided convention. Mid-p, doubled one-sided and alternative conditional orderings are not reported, and the sample odds ratio is not a confidence interval.

A worked example.

Frequently asked questions.

References& sources.

1. [1]R Core Team. R stats manual current 2026. fisher.test two-sided probability ordering. Retrieved 2026-08-06. independence: primary; access: open.