Analysis Syntax Editor - Keywords Tab
Select the Keywords tab of the General Classification and Regression Trees Syntax Editor or the General CHAID Analysis Syntax Editor to access options to enter command syntax to completely specify the analysis, as well as various parameters that are used in the computations. The available keywords can be typed directly into the Analysis syntax window or entered by clicking on the respective buttons on this tab to insert them in the Analysis syntax window at the location of the cursor. The unique keywords available for General Classification and Regression Trees are described below; note that the GC&RT analysis shares several keywords with the General Linear/Nonlinear Models (GLM) syntax; in particular, refer to GLM Syntax for details concerning how to specify designs.
GCHAID Syntax - Keyword BONFERRONI
BONFERRONI = {YES}
{NO}
Example. BONFERRONI = YES;
Optional keyword; specify YES to apply the Bonferroni adjustment to the p-value for splitting in CHAID; specify NO if such an adjustment is not required.
Applies to GCHAID
GC&RT and GCHAID Syntax - Keyword CatDV
CATDV = {NO}
{YES}
Example. CATDV = YES;
Required keyword; specify YES, if the response is categorical (for classification trees analyses), or NO if the response is continuous (for regression trees analyses).
Applies to GC&RT, GCHAID
GC&RT and GCHAID Syntax - Keyword CountV
COUNTV = { NONE}
{Variable}
Example. COUNTV = FREQUNCY;
Optional keyword; specify a count variable that indicates the number of actual observations in a case; NONE is default, meaning that each case represents one observation.
Applies to GC&RT, GCHAID
GCHAID Syntax - Keyword EXHAUST
EXHAUST = {NO}
{YES}
Example. EXHAUST = YES;
Optional keyword; specify YES for exhaustive CHAID analysis and NO for standard CHAID analysis. For details, see also Basic Tree-Building Algorithm: CHAID and Exhaustive CHAID.
Applies to. GCHAID
GC&RT Syntax - Keyword Goodness
GOODNESS = {GINI}
{CHISQUARE}
{GSQUARE}
Example. GOODNESS = GINI;
Optional keyword; specify the measure to be used as the Goodness-of-fit-measure. Note that this option only applies to classification-type analyses, i.e., when the dependent variable is categorical in nature (CATDV=YES). See the Introductory Overview and Computational Details for additional information.
Applies to GC&RT; applies to classification-type analyses only
ISE = Positive value
Example. ISE=1;
Optional keyword; default value is 1.0. Specify Standard error rule for v-fold cross-validation; see the Quick specs dialog - Validation tab; see also keyword VFOLD.
Applies to GC&RT
GC&RT Syntax - Keyword Fraction
FRACTION = Positive value less than 1.0
Example. FRACTION = .05
Optional keyword; specify the parameter Fraction of objects as described in the context of the Stopping parameter group box of the Quick specs dialog - Stopping tab.
Applies to GC&RT
GC&RT and GCHAID Syntax - Keyword MaxNodes
MAXNNODES = Positive integer value
Example. MAXNODES=1000;
Optional keyword; specify the maximum number of nodes to be used as the stopping criterion; see also the description of the options in the Stopping parameter group box of the Quick specs dialog box - Stopping tab.
Applies to GC&RT, GCHAID
GCHAID Syntax - Keyword MERGESPLIT
MERGESPLIT = {NO}
{YES}
Example. MERGESPLIT = NO;
Optional keyword for CHAID; see the General CHAID Models Quick specs dialog box - Advanced tab and Basic Tree-Building Algorithm: CHAID and Exhaustive CHAID for details.
Applies to GCHAID
GC&RT and GCHAID Syntax - Keyword MinSize
MINSIZE = Positive integer value
Example. MINSIZE = 5;
Optional keyword; specify the minimum number of cases for a node. This number is used as a splitting criterion, meaning that the splitting continues as long as the number of cases belonging to the node exceeds the specified number.
Applies to GC&RT, GCHAID
GC&RT and GCHAID Syntax - Keyword MisCost
MISCOST = {EQUAL}
{USERSPEC(value list)}
Example. MISCOST = USERSPEC(0 1 2 1.5 0 1 1 1 0);
Optional keyword; specify the Misclassification cost for the classification-type analysis. Refer also to the description of the Quick specs dialog box - Classification tab for details. If User spec. is specified, then include a list of values in parentheses to specify the respective misclassification costs. The values should correspond in sequence to the row-wise elements of the k (rows) x k (columns) misclassification matrix for a categorical dependent (class) variable with k categories, where the columns reference the observed classes, and the rows reference the predicted classes.
Applies to GC&RT, GCHAID; applies to classification-type analyses only
GC&RT Syntax - Keyword NSurrogates
NSURROGATES = Positive integer value
Example. NSURROTAGES = 2;
Optional keyword; specify the Number of surrogates for the C&RT analysis; see the description of the Quick specs dialog box - Advanced tab for details. The default is 0 (no surrogate splits).
Applies to GC&RT
GCHAID Syntax - Keyword PROBMERGE
PROBMERGE = Probability value
Example. PROBMERGE = 0.05;
Optional keyword; specify value of the probability-for-merging-categories in CHAID. For details, see also Basic Tree-Building Algorithm: CHAID and Exhaustive CHAID.
Applies to GCHAID
GCHAID Syntax - Keyword PROBSPLIT
PROBMERGE = Probability value
Example. PROBSPLIT = 0.05;
Optional keyword; specify value of the probability-for-splitting in CHAID. For details, see Basic Tree-Building Algorithm: CHAID and Exhaustive CHAID.
Applies to GCHAID
GC&RT Syntax - Keyword StopRule
STOPRULE = { MISCLASS }
{ VARIANCE }
{ DEVIANCE }
{ DIRECT }
Example. STOPRULE = misclass;
Optional keyword; select the stopping rule to be used for building the tree; see also Pruning and Selecting the "Right-Sized" Tree, and the description of the C&RT Quick specs dialog box - Stopping tab for details. You can select to prune on misclassification error (for classification problems only), variance (for regression problems only), deviance (for classification problems only), or you can select FACT-style direct stopping.
Applies to GC&RT
PRIOR = {ESTIMATED}
{EQUAL}
{USERSPEC(value list)}
Example. PRIOR= USERSPEC( 0.2 0.2 0.2 0.4);
Optional keyword; specify the prior classification probabilities to be used in the classification-type analyses (not applicable to regression-type analyses; see also the Introductory Overview). If User spec. is specified on the Quick specs dialog box - Classification tab, include inside parentheses a list prior probabilities for each class (group) for the categorical dependent variable.
Applies to GC&RT; applies to classification-type analyses only
GC&RT and GCHAID Syntax - Keyword VFold
VFOLD = {NO}
{YES}
Example. VFOLD = YES;
Optional keyword; specify whether or not v-fold cross-validations is to be performed for selecting the best tree. See the Introductory Overview and Computational Details for additional information.
Applies to GC&RT, GCHAID
GC&RT and GCHAID Syntax - Keyword VFoldSeed
VFOLDSEED = Positive integer value
Example. VFOLDSEED = 123;
Optional keyword; only applicable if VFOLD = YES; specify a seed for the random number generator.
Applies to GC&RT, GCHAID
GC&RT and GCHAID Syntax - Keyword VFoldValue
VFOLDVALUE = Positive integer value
Example. VFOLDSEED = 10;
Optional keyword; specify the value for the V-fold value for the V-fold cross-validations option (number of folds for v-fold cross-validation). Refer to the GC&RT Introductory Overview and Computational Details for details.
Applies to GC&RT, GCHAID