A continuity correction applied to a chi-squared contingency table test. Normally only applied for a 2 × 2 table, where the effect is greatest and by a happy coincidence the arithmetic required is the simplest. It involves diminishing the numerical size of the difference between observed and expected values by 0.5 before squaring, i.e. the chi-square statistic calculated is

${\displaystyle \sum \frac{{\left(\left|O-E\right|-0.5\right)}^2}{E}}.$