ABSTRACT
In Chapter 5, residual plots were used to investigate the assumptions underlying the linear model for the case of a single qualitative variable (factor). As discussed and briey demonstrated in Chapter 12, all of these residual plots are also applicable to a model with a single quantitative variable or variate (i.e. SLR), as well as to the more complex regression models such as those described in Chapters 14 to 18. However, in models with quantitative explanatory variables, the additional question arises as to whether a linear trend is a good description of the relationship, and several diagnostic plots can be used to investigate this. Concepts such as inuence and leverage can also help to examine the impact of individual points on the tted line, and cross-validation techniques can quantify the predictive power of the model. In this chapter, we introduce these concepts, and review and introduce some techniques for checking the t of regression models.
