PROC CLUSTER with METHOD=AVERAGE (average linkage) computes the distance between two clusters being merged. How is this distance defined under average linkage?
Select an answer to reveal the explanation.
Short Explanation
Here's the deal — average linkage defines the distance between two clusters as the average of all pairwise distances between members of the two clusters. You get a middle-ground measure that avoids single linkage's chaining problem and complete linkage's compactness bias.
Full Explanation
Average linkage (METHOD=AVERAGE in PROC CLUSTER) defines the between-cluster distance as the arithmetic mean of all pairwise distances between every member of one cluster and every member of the other cluster. This intermediate criterion avoids the chaining problem of single linkage and the compactness bias of complete linkage. The option defining the distance between the two closest members, one from each cluster, is single linkage (METHOD=SINGLE). The option defining the distance between the two farthest members, one from each cluster, is complete linkage (METHOD=COMPLETE). The option defining the distance between the centroids, or arithmetic means, of the two clusters is centroid linkage (METHOD=CENTROID), which uses cluster centroids rather than all pairwise member distances.