A1377
Title: A general model for partially observed functional data
Authors: M Carmen Aguilera-Morillo - Universitat Politecnica de Valencia (Spain) [presenting]
Maria Durban - Universidad Carlos III de Madrid (Spain)
Pavel Hernandez - Universidad Carlos III de Madrid (Spain)
Abstract: In functional data analysis, it is usual to assume that all sample data are fully observed within their domain. However, there are situations where the sample data (curves, surfaces) are partially observed, i.e., contain gaps or missing parts. To deal with the problem of statistical modeling of partially observed multidimensional functional data, a generalized additive scalar-on-function regression model is proposed. In order to control the smoothness of the functional coefficient, a p-spline penalty is added to the model estimation. The functional model is estimated thanks to the connection with the mixed effects model, where the smoothing parameters and the model coefficients are estimated directly. The performance of the proposed model is tested on a simulated and a real dataset of images of air pollution from India.
