There are two broad classiﬁcations of violations of assumptions about error: systematic and idiosyncratic. As we shall see, the systematic violations are often best identiﬁed visually by examining a variety of specialized graphs. Violations suﬃcient to make us doubt either the estimated model or to question the appropriateness of the probability values for the test statistics usually stand out in these specialized graphs. Our remedy, just as it was for nonindependence, is to transform the data so that we may use our full range of techniques for building models. Idiosyncratic violations arise from wild observations, commonly referred to as outliers, that for one reason or another are dramatically inconsistent with the other observations. The presence of outliers frequently causes violations of normality and homogeneous variance assumptions. Hence, we begin by considering methods for detecting and resolving the problems of idiosyncratic outliers and then turn to the visual methods for detecting assumption violations. As a practical matter, it is usually best to address the idiosyncratic outlier issues ﬁrst, and then consider possible systematic violations of the normality and homogeneous variance assumptions.