Finding hidden clusters in structured data can be hindered
by the presence of masking variables. If not detected,
masking variables are used to calculate the overall similarities between units,
and therefore the cluster attribution is more imprecise.
The algorithm q-vars implements an optimization method to find the variables
that most separate units between clusters. In this way, masking variables can be
discarded from the data frame and the clustering is more accurate.
Tests can be found in Benati et al.(2017)