Regression model for compositional data using a spatial econometric approach
- Autores: T Yoshida, M. Tsutsumi
- Localización: Proceedings of the 6th International Workshop on Compositional Data Analysis: Girona, 1-7 de juny de 2015 / coord. por Santiago Thió Fernández de Henestrosa, Josep Antoni Martín Fernández, 2015, ISBN 978-84-8458-451-3
- Idioma: inglés
- Enlaces
-
Resumen
In recent years, regression models for compositional data that consider spatial dependency have been discussed mainly in the field of geology. In compositional data analysis, most methods integrating spatial dependency have been geo-statistical approaches; research using spatial econometric approaches has been limited to a few recent studies. Furthermore, these studies’ methods for considering spatial dependency have not taken into account the effects of neighboring pixels, commonly used in the field of image processing. Thus, the present study extends these studies’ approaches to spatial dependency by presenting an empirical regression model for compositional data with a spatial econometric approach. To do so, we briefly review past work on relevant regression model, and then, propose a method under which spatial dependency is assumed to decrease with geographical distance. We then test the method through an empirical examination using land-use data with the characteristics of compositional data; this allows us to evaluate the proposed method in terms of predictive accuracy.
