作者: Michael Haber
DOI: 10.1080/01621459.1980.10477503
关键词:
摘要: Abstract This article considers the two-sided test of independence in a 2 × contingency table. The chi-squared approximation to exact significance Pearson's X is compared with four methods adjusting for continuity. Computations were carried out nearly 150,000 fourfold tables. It was found that both uncorrected and Yates correction may give misleading outcomes even moderate sample sizes expected values, whereas other three provide closer approximations. Although fixed marginal totals are assumed, these corrections also be applied tables random totals, provided one interested conditional significance.