ABSTRACT
In the previous chapter, we examined the use of the least squares criterion in multiple regression models that allow us to examine the relationship between one or more predictors when the outcome is continuous. In this chapter, we are introduced to logistic regression, which can also be used when the outcome is categorical and that allows model prediction. Logistic regression and discriminant analysis (which is discussed in an upcoming chapter) share similarities, and there can be confusion on when one is more appropriate than the other. Understanding that you may not be fully familiar with discriminant analysis, we’ll offer the condensed version of how the two procedures contrast. The assumptions of multivariate normality and equal variancecovariance matrices, which are required in discriminant analysis, do not hold for logistic regression. Thus, logistic regression is more robust than discriminant analysis when these assumptions are not met. Additionally, logistic regression is oftentimes less interpretatively challenging than discriminant analysis given that it falls within the regression family, more common to many researchers as compared to discriminant analysis.
