As new data collection methods become increasingly common, traditional techniques for data visualization and dimension reduction are becoming less adequate. This article proposes a surrogate-assisted sufficient dimension reduction (SDR) method for regression with a metric-valued response on Euclidean predictors. The method extends classical SDR methods and applies to responses on compact metric spaces. The method’s superior performance is demonstrated through simulation experiments. The article also provides an analysis of the distributions and functional trajectories of county-level COVID-19 transmission rates in the U.S. based on demographic characteristics.
Publication date: 23 Oct 2023
Project Page: https://arxiv.org/abs/2310.12402v1
Paper: https://arxiv.org/pdf/2310.12402
