Yates' continuity correction adjusts the chi-square statistic for 2×2 tables to correct for the fact that the discrete chi-square distribution is being approximated by a continuous chi-square distribution. It subtracts 0.5 from each |O − E| value before squaring, producing a more conservative (smaller) chi-square value and larger p-value.
The corrected chi-square for a 2×2 table:
Recommended when:
When expected counts are below 5, use Fisher's exact test instead.
Many statisticians consider Yates' correction overly conservative. It was proposed in 1934 when computing Fisher's exact test was difficult. Today, exact tests are easily computed, making Yates' correction less necessary. CrossTabs.com provides both corrected and uncorrected chi-square values plus Fisher's exact p-value so you can compare all three approaches.
No. Many modern statisticians recommend against routine use because it is overly conservative (inflates p-values). Use uncorrected chi-square when expected counts are ≥ 5, Fisher's exact test when they are < 5, and Yates' only if you specifically want a conservative result.
No. Yates' correction is only defined for 2×2 tables. For larger tables, the chi-square approximation is generally adequate when expected counts are ≥ 5.
The correction always reduces the chi-square value and increases the p-value. The effect is most noticeable with small samples. For large samples (N > 100), the difference is usually negligible.