callHierarchyTree {rchic} | R Documentation |
Reads the data, prepare the data and call hierarchyTree.
callHierarchyTree(fileName, contribution.supp = FALSE, typicality.supp = FALSE, computing.mode = 1, verbose = FALSE)
fileName |
name of the file containing the data |
contribution.supp |
boolean to compute the contribution of supplementary variables |
typicality.supp |
boolean to compute the typicality of supplementary variables |
computing.mode |
controls the computing mode: 1=classic implication, 2=classic implication+ confidence, 3=implifiance |
verbose |
boolean to give many details. |
This function allows you to compute the hierarchy tree. This tree is built with the cohesion index which is not symmetric.
It is very important to keep this in mind. At the end of the computation we obtain the hierarchy tree,
as follows. In this tree, couple of variables are gathered according to their cohesion measure. Significant levels
are highlighted in red. A significant level means that the considered level is more significant than the previous level
and than the next one. In the following figure and in all the cases, an expert must choose according to his knowledge
what is the threshold in the hierarchy tree after which the cohesion measure is not significant for the data considered.
In general the last classes in the hierarchy tree are not significant. Comparing the hierarchy tree with the similarity tree,
we can see that the hierarchy tree contains a greater number of classes than the similarity tree. As the cohesion is not
symmetric, we have an important information on the implication. In this example, we can observe that the strong class is:
craftiness generally implies malignant. The second strongest class is: powerful generally implies strong.
Raphael Couturier