Classification Trees Results - Tree Structure Tab

Computational Methods

Select the Tree structure tab of the Classification Trees Results dialog box to access the options described here.

Classification tree structure. Click the Classification tree structure button to display the Tree Structure spreadsheet, which displays all the information available to be produced on the Tree graph for the selected tree. This includes for each node in the selected tree:

  1. The left branch and right branch child nodes to which cases or objects are sent if they satisfy, or do not satisfy, respectively, the split condition at a split node.

  2. The number of cases or objects in each observed class that are sent to the node.

  3. The predicted class to which cases sent to the node are assigned.

  4. Information detailing the split condition for a split node. Note that no child nodes or split conditions are displayed for the terminal nodes of the tree.

Child nodes. Click the Child nodes button to display a spreadsheet that contains for each split node in the selected tree the left branch and right branch child nodes to which cases or objects are sent if they satisfy, or do not satisfy, respectively, the split condition at the node. Note that no child nodes are displayed for the terminal nodes of the tree.

Class compositions. Click the Class compositions button to display a spreadsheet that contains for each node in the selected tree the same information as in the Child nodes spreadsheet (see above), and additionally the number of cases or objects in each observed class that are sent to the node.

Predicted classes. Click the Predicted classes button to display a spreadsheet that contains for each node in the selected tree the same information as in the Class compositions spreadsheet (see above), and additionally the class to which cases or objects sent to the node are assigned.  

Split conditions. Click the Split conditions button to display a spreadsheet that contains information describing the split condition for each split node in the selected tree. If you selected either Discriminant-based univariate splits for categ. and ordered predictors or C&RT-style exhaustive search for univariate splits as the Split selection method for the analysis, this information includes the Split constant and the Split variable for the split condition. If the split variable is a categorical predictor, the Split category (or categories, if more than one) will be displayed, and if the split variable is an ordered predictor, the Split coefficient will be displayed. If Discriminant-based linear combination splits for ordered predictors has been selected as the Split selection method for the analysis, the Split constant and the Split coefficient for each ordered predictor variable will be displayed. Note that no split condition information is displayed for the terminal nodes of the tree.  

Tree sequence. Click the Tree sequence button to display a spreadsheet that contains the tree sequence with the following options considered:

  • If you select a Pruning Stopping rule for the analysis, the spreadsheet will contain the number of Terminal nodes, the CV cost and its standard error, the Resubstitution cost, and the Node complexity for each pruned tree. The pruned tree selected as the "right-sized" tree is marked with an asterisk.

  • If you select FACT-style direct stopping as the Stopping rule for the analysis, the same information is presented for the selected "right-sized" tree.

Cost sequence. Click the Cost sequence button to display the Cost Sequence line graph. If you selected a Pruning Stopping rule for the analysis, the graph plots the CV cost and the Resubstitution cost for each pruned tree. The pruned tree selected as the "right-sized" tree is marked with an asterisk. If you selected FACT-style direct stopping as the Stopping rule for the analysis, the same information is presented for the selected "right-sized" tree.

Predictor importance. Click the Predictor importance button to display a spreadsheet that contains the importance ranking on a 0-100 scale for each predictor variable in the analysis. This option is available when you have selected either Discriminant-based univariate splits for categ. and ordered predictors or C&RT-style exhaustive search for univariate splits as the Split selection method for the analysis. See also, Predictor Importance in STATISTICA GC&RT, Interactive Trees, and Boosted Trees.

Importance plot. Click the Importance plot button to display a bar graph that displays the importance ranking on a 0-100 scale for each predictor variable in the analysis, and gives a quick visual indication of the most important predictor variables. This option is available when you have selected either Discriminant-based univariate splits for categ. and ordered predictors or C&RT-style exhaustive search for univariate splits as the Split selection method for the analysis. See also, Predictor Importance in STATISTICA GC&RT, Interactive Trees, and Boosted Trees.