S. Navarro Díaz, J. F. Vera Vera
For the analysis of large dissimilarity data sets in MDS, cluster and multidimensional scaling techniques has been proposed in conjunction to lead a better understanding. Several models have been proposed to perform clustering of the objects while simultaneously the cluster centres are represented in a low dimensional space using MDS, in particular for one-way one-mode datasets. In this work, the precision with which the cluster centres are represented in this framework is analysed for the interpretation purpose. If no probabilistic hypotheses are assumed, resampling techniques are considered in conjunction with the corresponding block-shaped partition of the dissimilarity matrix, to determine confidence regions for the cluster centres representation in a cluster-multidimensional scaling framework. Several alternative methodologies to this end are analysed and artificial and real datasets are analysed to study the performance of the proposed procedure.
Palabras clave: Multidimensional Scaling, Resampling Techniques, Stability, Confidence regions, Cluster
Programado
X04 Pausa Café. Sesión Posters. Reunión TEST - Edificio 1
7 de septiembre de 2016 11:40
Edificio 1
