A1466
Title: Optimal experimental designs for low-rank function completion
Authors: MingHung Kao - Arizona State University (United States) [presenting]
Abstract: The focus is on optimal experimental designs for collecting high-quality sparse functional data, consisting of observations of one or more random functions X1(t), ..., XM(t) taken at sparse and possibly irregularly spaced points over a compact domain T. Common objectives in analyzing such data include recovering the functions Xm(t) across T, and predicting a response function Y(t) using X1(t), ..., XM(t) as predictors. Drawing on ideas from low-rank matrix completion, a low-rank function completion (LRFC) framework has recently been proposed to efficiently carry out such analyses. However, optimal design strategies to ensure precise inference under the LRFC framework have not been studied, and existing design approaches for sparse functional data analysis (FDA) may become unwieldy under this framework. An efficient method is proposed for identifying optimal designs tailored to the LRFC framework. Its effectiveness is demonstrated, and its applicability to studies where other sparse FDA methods are considered is highlighted.
