Programmer Guide/Command Reference/EVAL/haclust

haclust

Hierarchical-agglomerative cluster analysis.

Usage:

haclust(d, m)

haclust(h, c)

Parameters:

d

A dissimilarity or distance matrix. For N data points this is a NxN matrix (e.g. created using the EVAL function dist).

m

The cluster method:

0 Single Linkage

1 Complete Linkage

2 Weighted Average Linkage

3 Average Linkage

4 Median

5 Centroid

6 Ward

h

A cluster hierarchy.

c

The number of clusters.

Result:

When using the parameters d and m, the result is a cluster hierarchy. This is a 3x(N-1) table containing the merged cluster pairs and the agglomeration niveaus.

The parameters h and c cause haclust to create a partition, which is a 1xN table containing the group indices.

Examples:

A simple demonstration of haclust usage.

[macro haclust_example]
        // create bivariate input data
        #datatable := new table * 2 /p num:x num:y
        $#datatable[*,0] := '1;2;1;3;4.5'
        $#datatable[*,1] := '1;0.5;4;2;4'
        
        // compute distance matrix
        #disttable := eval dist($#datatable)

        // cluster analysis methods:
        // 0 (SL), 1 (CL), 2 (WAL), 3 (AL), 4 (M), 5 (C), 6 (W)
        #method := 6    // Ward

        // nr of clusters
        #nclusters := 2

        // compute hierarchy and partition
        #hierarchy := eval haclust($#disttable,$#method)
        #partition := eval haclust($#hierarchy,$#nclusters)

        // show results
        showitem $#hierarchy
        showitem $#partition
exit