A simple X-Y plot is useful to test the linearity
assumption. This is accomplished one
independent (regressor) variable at a time.
The displayed plot has the regression residuals
as the dependent variable. Select an
independent variable to plot on the independent axis (X). If a curvilinear function is depicted, then
consider either a nonlinear regression model or data transformation. For example, consider the generalized
residual plots presented below. Note that Figure (a) and Figure (f) depict desired relationships;
one against the independent variable X, and the other over time. In each case the residuals fall within a horizontal
band centered on zero.
