Residual Plots

Figure 1. Warnings
Warnings

The Linear Regression procedure will not produce residual plots for WLS models; however, by saving predicted values and residuals, you can create weighted residuals and predicted values and produce a scatterplot yourself.

  1. To create weighted predicted values, from the menus choose:

    Transform > Compute Variable...

    Figure 2. Compute Variable dialog box
    Compute Variable dialog box
  2. Type wgtpred as the target variable.
  3. Type pre_1*sqrt(wgt_1) as the numeric expression.
  4. Click OK.
  5. To compute weighted residuals, click the Dialog Recall tool and select Compute Variable:
    Figure 3. Compute dialog box
    Compute dialog box
  6. Type wgtresid as the target variable.
  7. Type res_1*sqrt(wgt_1) as the numeric expression.
  8. Click OK.
  9. To produce the residuals plot, click the Dialog Recall tool and select Chart Builder, then click Reset to clear prior selections.
    Figure 4. Chart Builder
    Chart Builder
  10. Select the Scatter/Dot gallery and choose Simple Scatter.
  11. Select wgtresid as the y variable and wgtpred as the x variable.
  12. Click OK.
Figure 5. Scatterplot of Residuals by Square Footage
Scatterplot of Residuals by Square Footage

There is no pattern in this plot. The weights have corrected for the heteroscedasticity.

Next