F. D. Miranda Huaynalaya, M. D. Ruiz Medina
A Bayesian maximum a posteriori (MAP) estimation is adopted in functional regression from correlated surfaces (see Ruiz-Medina and Miranda, 2022). As an alternative, spatial curve regression is introduced in the spatial functional spectral domain in this paper. The derived numerical results show a significant impact of the truncation, tapering and smoothing parameters in both approaches. The present talk investigates this impact through a simulation study under different scenarios by a cross-validation analysis. The complexity and high-dimensionality inherent to the infinite-dimensional Bayesian approach makes indispensable a suitable selection criterion. On the other hand, the results from the spatial functional spectral approach are quite sensitive to a suitable choice of the orthonormal basis for projection. Conclusions are drawn for future extensions of both methodologies to alternative estimation frameworks.
Keywords: MAP estimation, nonparametric estimation, spatial curve regression, spatial periodogram operator, spatial spectral density operator, surface regression.
Scheduled
GT06 Functional Data Analysis IV. Functional Time Series
June 9, 2022 10:10 AM
Grade Hall