Preprint #97-6

On the Comparison of Two Proportions Using a Linear Model Analysis of 0, 1 Scorings

by

C. Philip Cox

Scorings of 0 and 1, for `failure' and `success' respectively, have been used in the comparison of two proportions. A linear model and the associated analysis of variance are here used to compare five `chi-squared-like' test statistics for such situations. One finding shows why, caeteris paribus, the standard, Pearson, `chi-squared-statistic' is necessarily less powerful than at least one of three alternative statistics. A numerical example shows that differences of the order of 5% between the magnitudes of alternative test statistics can occur in practice.

Keywords:2x2 tables, Catanova statistic, categorical data, chi-squared statistic, linear model, two proportions.


Copies of preprints are available from the author upon request. Use the preprint number (located at the top of the page) and make the request directly to the author, Iowa State University, Department of Statistics, Snedecor Hall, Ames, IA 50011-1210.