Functional Convolution Models, Design and Domain Selection
by Giles Hooker
Abstract: We present a functional extension of the distributed lag model in time series analysis. The functional convolution model is also a restriction of the historical linear model of Malfait and Ramsay (2000) for functional response models and is closely related to hemodynamic response models employed in fMRI studies. Our model is estimated via an integrated squared error approach and novel functional block bootstrap techniques are developed to provide inference. Identifiability of the model without penalization is non-trivial and opens problems in the design of functional covariates. The model also motivates the problem of selecting the domain of interest in a functional covariate for which a change-point approach is developed.
For More Information: contact: Mihee Lee. email: email@example.com