The user selects an R mode or a Q mode type of analysis. R mode will produce a dendrogram of variables. The goal is to determine if there are
An R mode dendrogram for the Iris data with the matrix of correlation coefficients as the similarity matrix is given below.
The first piece of output gives a summary of the sequence of linking -- once a pair of variables are linked or fused they cannot be separated.
Variables 3 and 4 link at the highest level. At 0.83 variable 1 links with the pair 3-4 and at -0.25 variable 2 links with triplet 1-3-4. The output is duplicated below. The sequence of linking can be used to modify the computer printout to a more satisfactory form.
Q mode analysis is handled the same way. The difficult part is where to draw the line separating "individual entities" and clusters. In the example above you could treat 3 and 4 as a fused pair and 1 and 2 as individuals. Or, you could treat 3-4 and 1-3-4 as clusters and 2 as an individual. Within the same dendrogram be consistent as to where you draw the boundaries.
I would begin a Q mode anlaysis by doing an R mode. If I found three clusters in the R mode I would look for three in the Q mode. Keep in mind that you may have a sample(s) or variable(s) which are outliers and just don't seem to fit.
