Composition-on-function regression model for the analysis of spectroscopic data

Speaker: Mara Bernardi
Affiliation: MOX, Dipartimento di Matematica - Politecnico di Milano
When: Wednesday 17th March 2021
Time: 09:15:00
Abstract: Silicate glasses are widely present in natural volcanic rocks and terrestrial planets are largely covered by such products. The study of the spectral response of glasses is of great importance in the context of planetary investigation to better interpret available and future remotely sensed spectra. In this work, we consider the problem of retrieving the chemical composition of a silicate glass based on its spectral response. This problem poses challenges from a modelling perspective as it involves heterogeneous and complex data whose structure and properties should be appropriately accounted for. We propose a regression model with compositional response and functional predictor. The model allows to account for the geometry of the space of compositions and for the inherent continuous nature of the spectrum. Moreover, we propose a nonparametric framework for inference and prediction. The dataset analysed concerns the spectral response of different series of silicate glasses presenting a range of chemical compositions that covers most rocks on planet Earth. Contact: Laura Sangalli,