Fisher's Exact Test Calculator

This calculator runs Fisher's exact test on a 2x2 contingency table, computing the exact hypergeometric probability of the observed table (line 22) and summing every table at least as extreme in the same row and column totals to build the two-tailed p-value (lines 25-35). At the defaults (Cell A=8, Cell B=2, Cell C=3, Cell D=7), that comes to a two-tailed p-value of 0.0698 — not significant at α=0.05, even though the observed table alone has only a 3.2% chance under the null (Exact Probability of This Table). Odds Ratio and Relative Risk are computed from the same four cells (lines 45-50), each capped at 9,999 rather than reported as literal infinity when the opposing cell is 0.

One counter-intuitive result from nudging each input up and down: Cell B and Cell C read as having zero effect on every output, but that is an artifact of this calculator's small integer defaults (2 and 3) combined with rounding — `Math.round(inputs.cellB || 0)` (line 5) rounds a 10% move to 2.2 or 1.8 both back to 2, so the value the engine actually sees never changes, even though Cell B and Cell C are both very much part of the row and column totals every probability term depends on. At the defaults, Cell A and Cell D happen to move the two-tailed P-Value by exactly the same amount when each is nudged the same way — a genuine tie produced by this specific table's row and column totals, not a case where one of the two cells matters more than the other. What this does not account for: any correction for multiple comparisons, and it treats every 2x2 table symmetrically regardless of whether the underlying design was a randomized experiment or an observational study, where an odds ratio and a relative risk carry different interpretations.