An approach for the investigation and comparison of the data structure in the multidimensional space is proposed. It is based on three properties, namely, the direction of the data sets, the variance-covariance of the data points, and the location of the data sets' centroids. A number of tests have been studied and are presented. It is shown that the combined use of these parameters allows a satisfactory estimation of the representativity between two data sets. Simulated data, as well as real case studies are presented.