Small area estimation via Heteroskedastic Geographically Weighted Regression for functional data.

Small area estimation is studied under a Heteroskedastic Geographically Weighted Regression model for functional data. The calibrated spatio-functional model we propose assumes that the variance varies across the space, and that each local model (defined at each location) gives a local non parametric estimation of the variance. This approach improves the model performance in terms of predictive spatio-functional fit for small area estimation, as illustrated by a simulation study and financial data analysis.