GAIN Subcommand (TREE command)
The GAIN subcommand
controls the display of tables that summarize gain and index values
and other measures that indicate the relative performance of nodes
in the tree. You can obtain tables that summarize results by node,
percentile group, or both. Each keyword is followed by an equals
sign (=) and the value for that keyword.
Example
TREE risk BY income age creditscore
/GAIN SUMMARYTABLE=YES TYPE=[NODE, PTILE].
SUMMARYTABLE Keyword
The SUMMARYTABLE keyword controls whether a summary table of index values is shown.
- For scale dependent variables, the table includes node number, number of cases, and mean value of the dependent variable.
- For categorical dependent variables, the table includes node number, number of cases, average profit, and ROI (return on investment) values.
- For categorical dependent
variables, the summary table is not displayed unless profits have
been specified on the
PROFITSsubcommand.
YES. Display summary
table. This is the default for scale dependent variables
if the subcommand is omitted or SUMMARYTABLE=NO is not included on the subcommand.
NO. The summary table is not displayed.
CATEGORYTABLE Keyword
The CATEGORYTABLE keyword displays gain and index values for target categories of
the dependent variable.
- The dependent variable must be categorical.
- Target categories must be
specified on the
DEPCATEGORIESsubcommand. - A separate table is generated for each target category.
- If the dependent variable is scale or no target categories are specified, the keyword is ignored and a warning is issued.
YES. Display a table of gain statistics for each target category of the dependent variable. The table includes percentage gain, response percent, and index percent (lift) by node or percentile group. This is the default.
NO. No category table is shown.
TYPE Keyword
The TYPE keyword specifies the type of gain table(s) that are produced. The
keyword is followed by an equals sign (=) and one or both of the following
types enclosed in square brackets:
NODE. Terminal nodes. Rows of the table represent terminal nodes. This is the default.
PTILE. Percentile. Each row of the table summarizes a percentile group.
SORT Keyword
The SORT keyword specifies how rows are sorted in all gain tables.
DESCENDING. Rows are sorted in descending order of index value. This is the default.
ASCENDING. Rows are sorted in ascending order of index value.
INCREMENT Keyword
The INCREMENT keyword controls the width of percentile groupings for a percentile
gain table (TYPE=[PTILE]).
- The default increment is 10, which gives decile intervals.
- You can specify one of the following percentage increments: 1, 2, 5, 10, 20, or 25. If a different value is specified, a warning is issued and the default increment is used.
-
INCREMENTis ignored for node-by-node gain tables.
CUMULATIVE Keyword
The CUMULATIVE keyword specifies whether node-by-node tables (TYPE=[NODE]) include cumulative statistics. It is ignored
for percentile gain tables, which are always cumulative.
YES. Cumulative statistics are displayed. This is the default.
NO. Cumulative statistics are not displayed.