ABSTRACT
This chapter introduces basic concepts and describes standard hypothesis testing methods associated with least squares regression and Pearson's correlation. Both least squares regression and Pearson's correlation provide useful and important tools for describing and understanding associations. The chapter provides some indication of what might go wrong with these standard methods and how these concerns might be addressed. When classic methods are unsatisfactory, there is the issue of what might be done to correct known problems. The chapter also introduces some alternatives to least squares regression. There is a simple way of testing the hypothesis that Pearson's correlation is zero in a manner that allows heteroscedasticity. The following features of data influence the magnitude of Pearson's correlation: the slope of the line around which points are clustered; the magnitude of the residuals; outliers; restricting the range of the X values, which can cause r to go up or down; curvature.
