Wireless sensor networks (WSNs) are receiving an upsurge of research interest in both academia and industry. The key issue for the design and operation of WSNs is the optimization of power consumptions. Several approaches have been proposed to address this aspect and a very promising approach is known to be Â¿clusteringÂ¿, which foresees to allow only a subset of nodes in the network to send data (via compress and aggregate operations) to a common sink node (e.g., for data reporting in monitoring application). Recently, a novel clustering algorithm based on the concept of Â¿data similarityÂ¿ has been introduced and shown to provide good performance. In the present paper, we move from and generalize this latter clustering algorithm, as well as substantiate via computer simulations the advantages of our solution with respect to the original one. In particular, we extend the concept of data similarity from the perfect match of measured (i.e., raw) data to the statistical correlation of them. We also introduce the semi-variogram metric as a sound measure to estimate the statistical correlation among measured data. The novel algorithm is termed Data Similarity Variogram-based Clustering Algorithm (DSVCA), which will be proven to be a good solution for network's data traffic minimization and for reducing the energy consumptions of the overall network.