Working with the Residual Plot

A residual plot shows the relationship between the predicted value of an observation and the residual of an observation. The residual of an observation is the difference between the predicted response value and the actual response value. When using large data sets, the residual plot is displayed as a heat map instead of as an actual plot. In a heat map, the actual observations are binned, and the color of each point indicates the relative number of observations in that bin.

Residual plots have several uses when examining your model. First, obvious patterns in the residual plot indicate that the model might not fit the data. Second, residual plots can detect nonconstant variance in the input data when you plot the residuals against the predicted values. Nonconstant variance is evident when the relative spread of the residual values changes as the predicted values change. Third, in combination with other methods, the residual plot can help identify outliers in your data.

To change the residual that is plotted, click the icon on the vertical axis.

By default, a scatter plot or heat map is displayed. However, a histogram is also available. To display a histogram of the residuals, right-click in the Residual Plot, and select Use Histogram. In the histogram, you can change the residual that is plotted by clicking the icon on the horizontal axis.

From the histogram, it is easy to determine whether the distribution of the residuals is approximately normal or skewed. A non-normal residual histogram can indicate that the model does not fit the data.

The Residual Plot provides several features to investigate your data. These features are available in both the scatter plot and histogram. The process to activate these features is the same in either plot. The available features are plot the residuals by a category variable, show only the selected observations, exclude selected observations, and remove all exclusions.

There are two ways to select observations in the Residual Plot. If you hold the mouse over an individual point, you can click on that point to select it. Alternatively, when you click and drag the mouse in the Residual Plot, a selection rectangle appears. When you release the mouse, all points inside the rectangle are selected.

To plot the residuals by a category variable, right-click in the Residual Plot, select Plot By, and then select a category variable. A box plot of the residuals appears. The residuals are grouped by the selected category variable.

To show selected observations, select some data points, right-click in the Residual Plot, and select Show Selected. A table of the selected observations appears.

To exclude selected observations, select some data points, right-click in the Residual Plot, and select Exclude Selected. The selected observations are no longer used to build the model.

To remove all exclusions, right-click anywhere in the Residual Plot, and select Remove Exclusions. All observations are now used to build in the model.