Specifying the Criteria for Predictive Accuracy

Select the Predicted Classes tab of the Classification Trees Results dialog box to access the options described here.

Predicted class by observed class. Click the Predicted class by observed class button to display the Predicted Class x Observed Class n's spreadsheet. The spreadsheet contains the number of cases or objects in each observed class on the dependent variable (columns) classified as each class (rows). The Learning sample N is also displayed in the spreadsheet header. Click the accompanying 3-D histogram or Discrete contour plot button to display this information graphically, often making it easier to spot misclassifications when there is a large number of classes.

Assigned node by observed class. Click the Assigned node by observed class button to display the Terminal Node x Observed Class n's spreadsheet. This spreadsheet displays the number of cases or objects in each observed class on the dependent variable (columns) sent to each terminal node (rows). The Learning sample N is also displayed in the spreadsheet header. Click the accompanying 3-D histogram or Discrete contour plot button to display this information graphically, often making it easier to spot nodes that classify most of the cases in a class.

Prior
probabilities. Click the Prior
probabilities button to display the Class Prior Probabilities spreadsheet.
The spreadsheet header displays whether Prior probabilities were specified
to be Estimated, Equal,
or User specified via the Classification Trees Startup Panel - Methods tab. The Learning
sample N is also displayed. The specified a
priori probabilities for each class on the dependent variable are
displayed in the first column of the spreadsheet, and the n's for each
class are displayed in the second column.

Adjusted priors. Click the Adjusted priors button to display the Adjusted Prior Probabilities spreadsheet. The specified a priori probabilities for each class on the dependent variable, adjusted for the User-specified misclassification costs, are displayed in the spreadsheet. Note that this option is only available if you select User-specified misclassification costs for the analysis via the Classification Trees Startup Panel - Methods tab.

Class minimum objects. Click the Class minimum objects button to display the Class Minimum Objects for Stopping spreadsheet. The spreadsheet header displays the value entered in the Fraction of objects box on the Classification Trees Startup Panel - Stopping Options tab, and the spreadsheet displays the minimum N for stopping for each class on the dependent variable (i.e., the largest integer not exceeding the class N times the specified fraction of objects times the ratio of the class prior to the smallest class prior - using adjusted priors if you select User-specified misclassification costs for the analysis.). Note that this option is only available if FACT-style direct stopping is selected as the Stopping rule via the Classification Trees Startup Panel - Stopping Options tab.

Misclassification matrix. Click the Misclassification matrix button to display the Learning Sample Misclassification Matrix spreadsheet. The number of cases or objects in each observed class on the dependent variable (columns) misclassified as another class (rows) are displayed in the spreadsheet. The Learning sample N is also displayed in the spreadsheet header.

Misclassification costs. Click the Misclassification costs button to display the Misclassification Costs Matrix spreadsheet. The user-specified costs of misclassifying cases or objects in each observed class on the dependent variable (columns) as another class (rows) are displayed in the spreadsheet. Note that this option is only available when User-specified misclassification costs is selected for the analysis via the Classification Trees Startup Panel - Methods tab.