Hierarchical Clustering of histogram data using a "model data" based approach

Histogram data are usually used to represent complex phenomena for which is known not only the range of variability but even the inner variability. Several authors have proposed methods to analyze histogram data taking into account frequencies or density probability. In this paper we propose a different way to analyze histogram data. The idea is to take into account histogram shape. In doing that we propose to approximate histogram by a suitable mathematical model and to use model parameters to analyze phenomena described by means of histogram data. In particular, we will show how to transform histogram data in model data and subsequently how to do a cluster analysis on this data.