B0998
Title: Generalized functional additive mixed models with compositional covariates for Spanish Covid-19 incidence data
Authors: Sonja Greven - Humboldt University of Berlin (Germany) [presenting]
Matthias Eckardt - HU Berlin (Germany)
Jorge Mateu - University Jaume I (Spain)
Abstract: A generalized functional additive mixed model is extended to the situation when the outcomes are functions and parts of the independent variables are finite or infinite compositions, i.e. functional compositions, carrying relative information of a whole. Relying on the isometric isomorphism of the Bayes Hilbert space of probability densities and the space of square-integrable functions with integration-to-zero constraint through the centred log-ratio (clr) transformation, functional compositions are incorporated as functional covariates into the model using a flexible basis function representation that also accounts for the integration-to-zero constraint. The extended generalized functional additive mixed model allows for the estimation of linear, nonlinear and also time-varying effects of scalar and functional covariates, as well as of the effects of (potentially spatial) grouping factors, in addition to the compositional effects. The potential of the extended model is shown by estimating the effect of age curves, i.e. functional compositions, and smoking status on regional Covid-19 incidence data for Spain.
