ABSTRACT
This chapter considers methods for checking model assumptions and the use of transformations to correct problems with the assumptions. It begins by familiarizing readers with the look of plots that display no detectable patterns and deals with methods for checking the assumptions made in simple linear regression. The chapter presents methods for transforming the original data so that the assumptions become reasonable on the transformed data. Two of the error assumptions are independence and homoscedasticity of the variances. The main tool used in checking assumptions is plotting the residuals or, more commonly, the standardized residuals. Violations of the error assumptions are indicated by any systematic pattern in the residuals. This could be a pattern of increased variability as the predicted values increase, or some curved pattern in the residuals, or any change in the variability of the residuals. A handy way to identify cases with large leverages, residuals, standardized residuals, or standardized deleted residuals is with an index plot.
