Multiple Regression |

One of the assumptions made in carrying out hypothesis tests in regression analysis is that the errors are normally distributed (Myers 1986). You can use residuals to check assumptions about errors. For this example, the *studentized* residuals are used because they are somewhat better than ordinary residuals for assessing normality, especially in the presence of outliers (Weisberg 1985). You can create a distribution window to check the normality of the residuals, as described in Chapter 12, "Examining Distributions."

Choose Vars:Studentized Residual. |

A variable called **RT_GPA_1** is placed in the data window, as shown in Figure 14.20.

**Figure 14.20:** GPA Data Window with RT_GPA_1 Added

Notice the names of the last three variables. The number you see at the end of the variable names corresponds to the number of the fit window that generated the variables. See Chapter 39, "Fit Analyses," for detailed information about how generated variables are named.

Related Reading |
Linear Models, Residuals, Chapter 39. |

Copyright © 2007 by SAS Institute Inc., Cary, NC, USA. All rights reserved.