Title: Bayesian hierarchical functional mixed effects model with shape constrained Gaussian processes
Authors: Jangwon Lee - Korea University (Korea, South) [presenting]
Taeryon Choi - Korea University (Korea, South)
Abstract: A Bayesian hierarchical functional mixed-effects model for grouped data observed unequally spaced time is proposed. The method is formulated as a multivariate functional mixed-effects model whose mean part and random part modeled by a Bayesian spectral analysis with and without shape constrained, such as monotone, convex, U-shaped, and multiple-extremes. By assuming the hierarchical structure on the spectral coefficient, the model can capture an overall mean trend as well as group trend and subject-specific trend. For flexible modeling for serial dependence in the temporal data, we assume that the error term is a multivariate Ornstein-Uhlenbeck process. The inference is performed by Markov chain Monte Carlo methods.