| Abstract: | ![]()
In the article "The Use and Misuse of Chi-Square: Lewis and Burke Revisited," by Kevin L. Delucchi (Psychological Bulletin, 1983, Vol. 94, No. 1, pp. 166-176; see record 1984-00274-001), the standard error of gamma expressed in Equation 19 is incorrect. The variables under the radical should be reciprocals. The correct formula is published here. Delucchi's article reviewed the proper use of the Pearson chi-square for analyzing contingency tables. (The following abstract of the original article appeared in the record attached to the aforementioned citation): Reviews the proper use of the Pearson chi-square for analyzing contingency tables. The 1949 article by D. Lewis and C. J. Burke, who cited 9 sources of error in the use of chi-square, is updated. Since the publication of that article, statisticians have addressed the question of the minimal size of expected cell frequencies. This problem has been examined from 2 perspectives: tests of association hypotheses in contingency tables and testing goodness-of-fit hypotheses. Under certain conditions, expected cell frequencies less than 5 do not substantially alter the Type I error rate. Supplementary and alternative approaches to the chi-square, including those involving log- and log-linear models, log-likelihood ratio, partitioning, comparison of individual proportions, and analysis of ordered categories, are discussed. Emphasis is placed on techniques that are of use to the practicing researcher who often deals with qualitative ordered and unordered data. (PsycINFO Database Record (c) 2010 APA, all rights reserved) |
