Y. Ramirez, F. J. Rodríguez Cortés, J. Mateu, R. Restrepo
Principal components analysis (PCA) is a widely used statistical technique in the social and physical sciences. However in many cases this technique is applied without taking into account spatial dependence. The correlation that arise because the location of the observations can change drastically the analysis and its omission can lead the wrong interpretations. When PCA includes the spatial correlation, the analysis is known as geographical weighted principal component analysis (GWPCA), and in this case the explained variance needs to be corrected for a right interpretation. In this paper we consider GWPCA with a generalization of the Frobenius norm that provides an increase in the explained variance. We analyse two real data sets improving the results of previous analyses.
Palabras clave / Keywords: generalized Frobenius norm, geographical weighting, Laplacian, matrix decomposition, Principal components analysis, Singular value decomposition
Programado
Sesión V02 Estadística Espacial y Espacio-Temporal I
1 de junio de 2018 16:00
Sala 1
