produces decision trees (regression trees) for the specified variables. As the measurement level of a variable determines how a variable is treated, an initial dialogue asks you whether you wish to modify the corresponding property of your variables. Note that you can temporarily change the measurement level of a variable for this procedure using the contextual menu when selecting a variable in the Variables pane.

If the menu does not show the SPSS extension module "Decision Trees" has not been licensed.

First you will have to specify a dependent variable and the independent variables to be considered for inclusion in the tree. If the dependent variable is categorical, you can limit the analysis to specific categories or mark them as central ("target of primary interest") in your analysis (Categories). For nominal dependent variables user missing categories can also be included.

Growing method lets you choose one of three tree growing methods (defaults to CHAID) and optionally you can force the first variable to be included for the first split and specify an influence (weight) variable, Note that the several options briefly explained below vary according to the measurement level of the variables, as well as the tree growing method chosen.

• Output to be produced: tree is presented (orientation, size) and what statistics are to be displayed with the nodes. (This can also be changed later in the tree editor); other statistics to be displayed).
• Validation Validation options Validation separates the sample data set into training samples and testing samples, especially useful with large data sets (the tree building process requires intensive computing) : the tree is built using a (smaller) training sample and in a second step is validated using either the full data set or additional samples.
• Criteria for the tree growing method (minimal number of observations in parent/child nodes, statistical criteria etc.)
• Save Copy derived variables to the current dataset.

  Copy the following variables to the data matrix for further analysis: Terminal node number, predicted category (categorical variables) or value (scale variables) and predicted probability (for each category of the dependent variable (not available for interval scaled dependent variable).

• Options Missing values handling (depending on the tree growing method); more advanced options.

The tree will be shown in the output window, with statistics or charts. A double-click on the tree opens the Tree Editor, a tool that lets you inspect the tree in detail and change its appearance, e.g. orientation, type of information shown for each node (summary, chart or both), colours etc.

A tree map (a clickable mini-view of the tree, shown on the screenshot) lets you easily navigate larger trees. ()

Here's a screenshot of an open tree editor explaining the main functions.

• Properties Change fonts, color etc. for the selected nodes. (Options also available from the toolbar
• Open/Close summary window shows numerical or graphical details for selected nodes. Very useful if you select more than one node, as it shows the details for the selected nodes combined.

Besides studying the tree, it is often useful to examine the groups (nodes) defined by the tree building process in some more detail. has options to (applies to all selected nodes)

• defines and applies a to the current data set.
• writes SPSS commands (Filter) that let you select the corresponding observations for additional analysis to a command file for further uses. Alternatively you can also create derived variables for the observations in the selected nodes ("scoring").

See the PDF manual or on-line help for details for the many options. To produce a tree using default values you could simply write:

TREE lrscale [s] BY agea [s] gndr [n] eduyrs[s] polintr[o] prtdgcl[o].

The measurement level of a variable must be specified in square brackets: [s] for an interval scaled variable, [o] for an ordinal and [n] for a nominal variable.

• Hierarchical clustering [CLUSTER]
• K-Means Clustering [Quick Cluster]
• Two step cluster