ABSTRACT
Assessing the relationship between a predictor variable and a dependent variable is an essential task in the model-building process. If the relationship is identied and tractable, then the predictor variable is reexpressed to reect the uncovered relationship and consequently tested for inclusion into the model. Most methods of variable assessment are based on the correlation coefcient, which is often misused. The linearity assumption of the correlation coefcient is frequently not subjected to testing, in which case the utility of the coefcient is unknown. The purpose of this chapter is twofold: to present (1) the smoothed scatterplot as an easy and effective data mining method and (2) a general association nonparametric test for assessing the relationship between two variables. The intent of point 1 is to embolden the data analyst to test the linearity assumption to ensure the proper use of the correlation coefcient. The intent of point 2 is an effectual data mining method for assessing the indicative message of the smoothed scatterplot.
