Check Model Assumptions
For all models, it's easy to create residual plots to check model assumptions. Create residual-by-predicted plots to diagnose nonconstant error variance and identify outliers.
Residual versus Predicted Plot
Examine the distribution of residuals with quantile-quantile plots. Use partial leverage plots to examine changes in the residuals for the model with and without each explanatory variable.
Partial Leverage Plots
You can also create residual and predicted variables, partial leverage variables, and influence diagnostic variables. Influence diagnostic variables include hat diagonal, standardized and studentized residuals, Cook's D, Dffits, Covratio, and Dfbetas. You can plot any of these variables and use them to detect influential observations.
Cook's D versus Predicted Plot
Statistics and Operations Research Home Page | SAS/INSIGHT Software