CMStatistics 2017: Start Registration
View Submission - CMStatistics
Title: Prediction based on multivariate spatial data: A sufficient dimension reduction approach Authors:  Liliana Forzani - Universidad Nacional del Litoral (Argentina)
Maria Antonella Gieco - CONICET - Facultad de Ingenieria Quimica, UNL (Argentina)
Pamela Llop - Facultad de Ingenieria Quimica, UNL-CONICET (Argentina) [presenting]
Anne Francoise Yao - Universite Clermont Auvergne/LMBP (France)
Abstract: One of the principal objectives in spatial statistics is to reconstruct certain phenomenon of interest, using data measured over its region domain. The best known methods to perform such reconstruction are based mainly in predictions that use not only the variable of interest (response) but also extra variables (predictors) which are added to the model in order to get better results. When the amount of extra variables is large, it may be of interest to reduce it in order to simplify the analysis but without losing information about the phenomenon to be studied. The sufficient dimension reduction techniques (SDR) consists in reducing the high dimensional space of predictors combining them in a new set of variables that lives in an lower dimensional space without losing information about the response. We apply SDR techniques in order to perform prediction based on multivariate spatially correlated data. The good performance of our method is shown via some simulation studies and application to real examples.