A clusterwise regression strategy for spatio-functional data

In several geostatistical applications, collected data are curves spatially located. This implies that statistical techniques have to take into account the functional component of data and the relations due to their spatial location. In this paper, we propose a clusterwise linear regression strategy where the aim is to find simultaneously an optimal partition of the data and a set of functional regression models associated to each cluster. Such models take into account both the interactions among variables and the spatial relations among the observations.