Testing for a significance in spatial regression with functional response

Veronika Rímalová
Palacký University Olomouc, Czech Republic - Department of Mathematical Analysis and Applications
Tuesday 5th December 2017
Aula Saleri VI Piano - Edificio 14, Dipartimento di Matematica POLITECNICO DI MILANO
In this talk we present a problem consisting of modelling and testing for spatial functional data in the context of regression analysis. The motivating data set includes monthly measurements of 5 chemical elements taken from 5 different sites near Brno, Czech Republic. Sampling locations were selected with the purpose of testing if the sites can be divided in two parts, agricultural and forest soil, according to their chemical properties. We propose to treat the measurements taken from each sampling point as functions of time distributed in space. To test for the difference according to the soil type, we develop an interval-wise testing procedure, in a geostatistical framework. We propose to perform the tests on residuals of a spatial linear model with a functional response. We apply the proposed method to the available data and discuss the results from the geological viewpoint. (joint work with A. Menafoglio, A. Pini, E. Fišerová) Contact: alessandra.menafoglio@polimi.it