A0876
Title: Generalized linear model for spherical response with projection-based inference
Authors: Kipoong Kim - Changwon National University (Korea, South) [presenting]
Joern Schulz - University of Stavanger (Norway)
Sungkyu Jung - Seoul National University (Korea, South)
Abstract: Spherical observations arise in fields from medical imaging to microbiome research, yet regression models for spherical responses remain underdeveloped. Many existing proposals employ highly parameterized models to gain flexibility, making theoretical study challenging. To fill this gap, a framework based on a generalized linear model with von Mises-Fisher responses is introduced. Using standard likelihood theory under mild conditions, it is demonstrated that the maximum likelihood estimators are consistent and asymptotically normal. The model distinguishes covariate effects on mean direction from effects on dispersion and provides corresponding asymptotic inference. In simulation studies for a two-sample testing problem, the approach achieves higher robustness and power than conventional tests. Its practical utility is also demonstrated on a diffusion tensor imaging dataset, where it detects meaningful location and dispersion patterns.
