the hierarchical clustering serves as a labelling mechanism, in other words emulating  post-process inspection. Such inspection can be done either based upon one or more known features as recorded during the process, or on other properties such as taste, colouring, general emerging chemical electrical, chemical and other emerging physical properties not recorded during the manufacturing process. In this latter case the assumption is that the bad, rejected, cases will represent the extreme clusters in a hierarchical cluster structure. Therefore, we used this for the labelling in the paper. The dendrogram is merely there to show such a structure.

You may follow this papers........

http://84.89.132.1/~michael/stanford/maeb7.pdf

http://www.ncss.com/wp-content/themes/ncss/pdf/Procedures/NCSS/Hierarchical_Clustering-Dendrograms.pdf

.

you'll find a simple introduction to dendrogram in the context of gene expression data here :

http://compbio.uthsc.edu/microarray/lecture1.htm

.

in choosing a specific "algorism" for providing an intuitively elegant dendrographic classification of non genetic profiles (low back pain syndromes we found that a cluster analysis of clustering techniques favoured "iterative relocate shape" (in Clustan) though I suspect this would not be so suitable for genetically mediated mechanisms.

The agglomerative hierarchical clustering algorithms provides cluster hierarchy for acceptance of a specific result that is commonly displayed as a tree diagram called a dendrogram. You can find some reference from the attached papers.