McNemar Paired Binary Test Calculator
Analyze discordant outcomes in a matched binary table with exact or corrected approximate inference.
Loading tool…
How to use McNemar Paired Binary Test Calculator
Only discordant before/after counts b,c determine McNemar inference. Exact mode doubles the smaller Binomial(b+c,.5) tail, capped at1; corrected chi-square uses max(|b−c|−1,0)²/(b+c).
- Enter before yes / after yes, before yes / after no, before no / after yes, before no / after no.
- Choose the stated convention and check the units and supported model.
- Calculate, inspect the result, then copy or download it.
Good to know
- Whole paired counts, at most1000discordant pairs. Corrected chi-square p is withheld below25discordant pairs; no discordance gives a no-information state. Not an independent A/B test.
- Runs locally. Text/JSON fields are capped at30,000 characters; stricter page-specific and numeric-string limits also apply. Numeric inputs are bounded to magnitude1e15 and nonzero magnitude at least1e-100 unless an exact-string mode states otherwise. Approximate summaries use12 significant digits; CSV numeric cells retain the complete stored value and potentially formula-like string cells gain a protective apostrophe; unsupported, non-finite or detected underflow states return an error.
Common questions
Do the two concordant cells affect the test statistic?
No. They contribute to the total pair count, but only pairs that changed outcome enter the stated discordance test.
Is the input uploaded or saved?
No. The tool calculates inside this browser tab and provides a local download; it does not persist your input.
Keep going with related tools
All math →Binomial Probability Calculator
Find exact-count and cumulative probabilities for independent trials with a fixed success chance.
MathTwo-proportion A/B Test Calculator
Compare two independent success rates with an explicitly approximate two-sided pooled z test.
MathChi-square Independence Calculator
Inspect a two-way count table with Pearson’s statistic, expected counts, and a conditional approximate p value.