CRITERIA Subcommand (QUICK CLUSTER command)

CRITERIA specifies the number of clusters to form and controls options for the clustering algorithm. You can use any or all of the keywords below.

• The NOINITIAL option followed by the remaining steps of the default QUICK CLUSTER algorithm makes QUICK CLUSTER equivalent to MacQueen’s n-means clustering method.

CLUSTER(n). Number of clusters. QUICK CLUSTER assigns cases to n clusters. The default is 2.

NOINITIAL. No initial cluster center selection. By default, initial cluster centers are formed by choosing one case (with valid data for the clustering variables) for each cluster requested. The initial selection requires a pass through the data to ensure that the centers are well separated from one another. If NOINITIAL is specified, QUICK CLUSTER selects the first n cases without missing values as initial cluster centers.

MXITER(n). Maximum number of iterations for updating cluster centers. The default is 10. Iteration stops when the maximum number of iterations has been reached. MXITER is ignored when METHOD=CLASSIFY.

CONVERGE(n). Convergence criterion controlling minimum change in cluster centers. The default value for n is 0. The minimum change value equals the convergence value (n) times the minimum distance between initial centers. Iteration stops when the largest change of any cluster center is less than or equal to the minimum change value. CONVERGE is ignored when METHOD=CLASSIFY.