B0994
Title: Experimental designs for functional data analysis
Authors: MingHung Kao - Arizona State University (United States) [presenting]
Abstract: Functional data analysis (FDA) has gained much popularity in extracting useful information from repeated measurements collected at various points in a domain, such as time. A crucial step for rendering a precise and valid inference is to have a high-quality sampling schedule to sample informative data from the underlying random function. We are concerned with this design problem for FDA, and propose efficient computational approaches for obtaining good designs to rein in cost. Our proposed approach generates high-quality designs to allow a precise recovery of the underlying function, as well as precise prediction with functional regressions.