Title: Directed local testing in functional linear regression
Authors: Christoph Rust - University of Regensburg (Germany) [presenting]
Abstract: A directed testing procedure is developed for the coefficient function of the Functional Linear Model (FLM) with a scalar response. The suggested procedure provides local interpretability of the regression function in the sense that it allows us to infer from the data a threshold up to which the random curves have a significant influence on the scalar outcome variable. The test is applicable in situations where one can assume that the functional predictors influence the outcome variable only on a sub-interval of the domain. Such situations regularly occur for analyzing brain-image data but also in regional economics. Theoretical results are established, and their finite sample performance is assessed by means of a simulation study and shows applicability of the suggested testing procedure with a real data application.