New index for clustering tendency Original Research Article

A new index for clustering tendency is described. The index is based on the frequency distribution of the lengths of the edges in the minimum spanning tree connecting the objects, compared with the probability distribution of the lengths of edges of the minimum spanning tree connecting the same number of objects described by variables extracted from the uniform distribution. The here suggested index shows some advantages when compared with the Hopkins original index and with its modification suggested by Fernández Pierna and Massart. It can be used both to detect clusters, to measure the degree of non-uniformity of a data set (as required in many cases of multivariate calibration and QSAR studies), and to detect outliers.