Working with the Tree Window

The Tree window contains the decision tree, tree overview, and icicle plot.

Tip

To navigate the decision tree, you can use the mouse and keyboard. Hold down the Shift key and click anywhere in the Tree window to move the decision tree within the window. Use your mouse’s scroll wheel to zoom in and out of the decision tree. Scroll up to zoom in, and scroll down to zoom out. The zoom is centered on the position of your cursor.

The color of the node in the icicle plot indicates the predicted level for that node. When you select a node in either the decision tree or the icicle plot, the corresponding node is selected in the other location. When you select a leaf node, that node is selected in the Leaf Statistics window. A legend is available at the bottom of the model pane.

When the response variable is a measure variable, a gradient is used to denote the predicted bin. Darker colors represent larger values.

Right-click outside of a node in the Tree window to open a pop-up menu. The first item in this menu is Derive a Leaf ID Variable. When you click this item, SAS Visual Statistics creates a category variable that contains the leaf ID for each observation. You can use this variable as an effect in other models.

Right-click inside a node to open a different pop-up menu. The available menu options depend on whether you clicked a leaf node.

For leaf nodes, you can select from the following menu options:

Split

opens the Split Decision Tree window. Use this window to select the variable that is used to split the node. Click OK to split the node based on the selected variable. Click Cancel to not split the node. Variables are sorted in descending order by their log worth.

Some variables are not available for a split if the value of the split is too small or the split would violate the Leaf size property.

Split Best

splits the node based on the variable with the best information gain ratio when Rapid growth is enabled. In addition, splits the node based on the variable with the best information gain when Rapid growth is disabled.

Train

opens the Train Decision Tree window. Use this window to train more than one level beyond the leaf node. First, select every variable that you want to be available for training. Only those variables selected in the Train Decision Tree window are available for training. Specify the maximum depth of training in the Maximum depth of subtree property. Click OK to train the decision tree.

For other nodes, select Prune to remove all nodes that follow the selected node. This turns the selected node into a leaf node. After pruning a node, you can select Restore to undo the prune.