View Article - PDF     Dataset (text)

Title: On The Use of Indicator Variables in Regression Analysis
Author: Keith M. Bower
Publication Source: International Society of Six Sigma Professionals, EXTRAOrdinary Sense, Vol. 2 No. 6, November 2001, pp. 1, 3

Abstract

An example illustrates the need to use indicator variables when performing regression analysis using one or more categorical predictor variables.

Notes

This paper was to address an issue I had seen several times with various organizations, especially those deal with day-of-week information. In particular, individuals had coded Monday = 1, Tuesday = 2, , Sunday = 7 and used simple linear regression of the form Y = β₀ + β₁*Day + ε. My guess is that they had used this approach when using ANOVA and as it worked there decided that it would also be appropriate here. It is not, of course.

The aim of this paper was to show by way of an example how misleading results may be obtained, and to show the need for building an appropriate model using indicator variables. I kept the argument to just three levels (three locations) to make the argument relatively straightforward and keep the statistical output to a minimum, but still be able to put the point across.

To download Adobe Acrobat (for free) click here.