Transformations of the Response

In Chapter 5, we discussed graphical diagnostic tools for inspecting residuals and checking the assumptions underlying the linear model (see Section 4.1). In particular, we considered how to evaluate the assumptions that the deviations have a common variance (homogeneity of variances, Section 5.2.1) and come from a Normal distribution (Section 5.2.3). We indicated that transformations of the response variable might be used to remove skewness in the distribution of the residuals (i.e. evidence of a non-Normal distribution) or to correct systematic changes in the variances of the residuals (i.e. evidence of variance heterogeneity, as illustrated in Figure 5.3).