Scholion.ai

An Analysis of Variance for Categorical Data, II: Small Sample Comparisons with Chi Square and other Competitors

Barry H. Margolin, Richard J. Light

Journal of the American Statistical Association · 1974 · Volume 69 · Issue 347

DOI: 10.1080/01621459.1974.10480201

Abstract

Exact small sample behavior in two-way contingency tables is investigated for Pearson's chi-square statistic ( X 2 ), Light and Margolin's C statistic and its related R 2 measure of association, Kullback's minimum discrimination information statistic (2Î), and Goodman and Kruskal's Lambda. R 2 is shown to be identical to Goodman and Kruskal's t b , leading to a test for independence based on t b . In small samples from a product of multinomials model, the null distribution of C is better approximated by a χ 2 distribution than is the null distribution of X 2 ; both are considerably better approximated by a χ 2 distribution than is the null distribution of 2Î. It is proved for tables with two columns and any number of rows that if the column totals are equal, then X 2 ≤ 2Î; thus, X 2 is more conservative than 2Î. Hence, use of 2Î should be avoided in testing independence in tables with small samples.

Enable JavaScript for citation tools and additional paper interactions.