MerQuR+Statistics Glossary

Statistics Glossary → Dendrogram and cophenetic correlation

Dendrogram and cophenetic correlationconcept

A tree diagram showing in which order, and at what distance, observations (or variables) are merged in hierarchical clustering; the cophenetic correlation measures how faithfully this tree reflects the original distances.

Agglomerative hierarchical clustering starts with each observation as its own cluster and at each step merges the two closest clusters. In a dendrogram the horizontal axis shows the observations and the vertical axis the distance (height) at which each merge happens. When the tree is cut horizontally at a given height, the branches below the cut form the clusters; long vertical branches point to clusters that are clearly separated from each other.

The cophenetic distance between two observations is the height at which they first join the same cluster. The cophenetic correlation is the correlation between these distances and the original distances over all pairs of observations; a value close to 1 shows that the tree preserves the distance structure of the data well. Different linkage methods (single, complete, average, Ward) can produce very different trees from the same data, and the cophenetic correlation helps compare them. Observations that sit side by side in a dendrogram are not necessarily close; closeness is shown by the merge height, not the left-to-right order.

An everyday example

A linguist clusters ten languages by vocabulary similarity. In the dendrogram, Spanish and Portuguese join at a very low height and then merge with Italian, while German and Dutch meet on a separate branch, and the two branches join only near the top. Cutting the tree at a height between the two branches separates the Romance and Germanic languages into two clusters.

In MerQur

In Hierarchical Clustering, the 'Chart' and 'Dendrogram cut' options draw the dendrogram, and the 'K-selection criteria table' also reports the cophenetic correlation. In Variable Clustering, 'Additional output: Dendrogram and k scan' produces an average-linkage dendrogram based on the 1 - |r| distance.

Analyses where this concept appears

Variable Clustering (VARCLUS)Hierarchical Clustering

Related concepts

Linkage method (Ward, complete, average, single)