CMStatistics 2022: Start Registration
View Submission - CMStatistics
B1888
Title: Functional mediation analysis with model selection Authors:  Yuko Araki - Tohoku University (Japan) [presenting]
Abstract: Mediation analysis has been developed when data are curves or images. We first estimate a set of functions to represent data using basis expansions. We have selected which basis function should be used among several candidate functions and how many basis functions should be used. For the mediation analysis model, a few types of functional regression models are used to represent the direct and indirect effects. Further, the proposed model selection criterion provides a powerful approach to identifying causal relationships. We conducted a simulation study and real data examples to examine the performance of the proposed model.