Caution on Statistics
1 Prevention Research Program, National Institute of Child Health and Human Development, Executive Plaza North, Rm 640, Bethesda, MD 20892
To the Editor.
We read with some concern the article by Maisels and Gifford.1 Although we agree with the results as they are calculated, we would like to express our concern about the logistic modeling used in the analysis. Most important, we would like to caution prospective users of logistic regression of the hazards of using the stepwise procedure. Although, the stepwise procedure will win you variables that demonstrate significant associations, it may exclude variables that are more important biologically to modeling the risk factors for the disease being studied.2




