GROWTHLIMIT Subcommand (TREE command)
The GROWTHLIMIT
subcommand specifies stopping criteria that limit the size of the
tree. Each keyword is followed by an equals sign (=) and the value
for that keyword.
Example
TREE risk [o] BY income age creditscore
/METHOD TYPE=CRT
/GROWTHLIMIT MAXDEPTH=4 MINCHILDSIZE=10.
MAXDEPTH Keyword
MAXDEPTH
specifies the maximum number of levels of growth beneath the root
node. You can change the maximum depth to adjust the size of the
tree.
AUTO. Three levels for CHAID and Exhaustive CHAID, five levels for CRT and QUEST. This is the default.
value. User-specified value. The value must be a positive integer.
MINPARENTSIZE Keyword
MINPARENTSIZE
specifies the minimum number of cases required to split a node.
Nodes with fewer cases are not split. You can use this setting to
avoid splitting nodes that have few cases.
- The size value must be a positive integer. The default is 100.
-
MINPARENTSIZE
must be greater thanMINCHILDSIZE
.
MINCHILDSIZE Keyword
MINCHILDSIZE
specifies the minimum number of cases in any child node. A node
will not be split if any of the resulting child nodes would have fewer
cases than the specified value.
- The size value must be a positive integer. The default is 50.
-
MINCHILDSIZE
must be less thanMINPARENTSIZE
.
MINIMPROVEMENT Keyword
For CRT, you can use MINIMPROVEMENT
to specify the minimum decrease in impurity.
The CRT growing method attempts to maximize within-node homogeneity.
In other words, a terminal node in which all cases have the same
value for the dependent variable is a homogeneous, "pure" node. A
node is not split if impurity would decrease less than the specified
value.
The improvement value must be positive value. The default is 0.0001. As the value increases, the number of nodes in the tree tends to decrease.