X hits on this document

Word document

Dichotomizing Continuous Variables: A Bad Idea - page 6 / 6





6 / 6

because of my ignorance, but because I have learned that my audience typically cannot understand interactions between continuous variables.  For example, I found that ethical idealism (continuous) moderated the relationship between misanthropy and support for animal rights, but I then dichotomized idealism for the analysis that I presented.

So, when can EGA be justified?  In exploratory research, especially where data are expensive, EGA may be justified as a method for determining whether or not there exists any relationship between X and Y (and hopefully not a quadratic one).  Dichotomization of X should not generally be part of EGA, but may be a last resort transformation to meet the normality assumption of the correlation analysis used to relate X to Y -- but other transformations and analyses may well be superior.  When one’s primary interest is in demonstrating an interaction, one may deliberately oversample extreme scores, which should increase power, but will also overestimate the size of the interaction effect in the population of interest.


Irwin, J.R., & McClelland, G.H. (2003).  Negative consequences of dichotomizing continuous predictor variables.  Journal of Marketing Research, 40, 366-371.

MacCallum, R.C., Zhang, S., Preacher, K.J., & Rucker, D.D. (2002). On the practice of dichotomization of quantitative variables. Psychological Methods, 7, 19-40.

Preacher, K. J., Rucker, D. D., MacCallum, R. C., & Nicewander, W. A. (2005). Use of the extreme groups approach: A critical reexamination and new recommendations. Psychological Methods, 10, 178-192.

Return to Wuensch's Stat Help Page

Karl L. Wuensch, Dept. of Psychology, East Carolina Univ., Greenville, NC  USA November, 2006.

Document info
Document views21
Page views21
Page last viewedFri Jan 20 00:22:52 UTC 2017