A1213
Title: An individually optimizable smoothing spline GMANOVA model
Authors: Hirokazu Yanagihara - Osaka Metropolitan University (Japan)
Mariko Yamamura - Otemon Gakuin University (Japan) [presenting]
Abstract: In situations where continuous response variables are repeatedly measured, the GMANOVA model provides a statistical framework for characterizing the effects of explanatory variables as time-varying processes. To flexibly estimate temporal changes in regression coefficients, smoothing splines are often employed. However, when multiple explanatory variables are included in the model, it is necessary to specify multiple smoothing parameters corresponding to the respective varying coefficients. The simultaneous optimization of these parameters is often computationally challenging in practical data analysis, leading to a substantial computational burden and instability in estimation. To address this difficulty, a smoothing spline GMANOVA model is proposed in which an orthogonal transformation is applied to the estimation framework. This transformation effectively decouples the varying coefficients, allowing each smoothing parameter to be optimized individually rather than simultaneously. As a result, the proposed approach alleviates the practical limitations associated with conventional models and provides a more stable, computationally feasible framework for analyzing longitudinal data.
