August 1978

Incorrect Use of χ2

Author Affiliations

Department of Medical Computer Science Southwestern Medical School The University of Texas Health Science Center at Dallas 5323 Harry Hines Blvd Dallas, TX 75235

Am J Dis Child. 1978;132(8):819. doi:10.1001/archpedi.1978.02120330091024

To the Editor.—An article by Weston et al (Am J Dis Child 131:1241-1242, 1977) exemplifies a continuing misuse of χ2 methodology in the literature. This is the use of χ2 to compare things other than the number of "successes" in N trials.1(p43)

For the χ2 method to be proper, the mean percent of cells responding would have to include the total number of cells used in each group. What the authors wanted to do was compare two groups of data that just happened to be percent cells responding for each subject.

The most appropriate method of analysis would have been the Mann-Whitney U test.1(p109) The usual t test could have been used but it would have required an arc sine transformation of each value,1(p185) a tedious problem.

Weston et al found no substantial difference between newborns and controls for monocyte phagocytosis and bactericidal activity. Using the summary statistics provided, I

First Page Preview View Large
First page PDF preview
First page PDF preview