Title: Functional clustering and missing value imputation of longitudinal data
Authors: Pai-Ling Li - Tamkang University (Taiwan) [presenting]
Abstract: A functional data approach is proposed for clustering and missing value imputation for incomplete longitudinal data. We adopt the notion of subspace-projected functional data clustering that each observed trajectory is viewed as a realization of a random function and is drawn from a mixture of stochastic processes, where each subprocess represents a cluster with a cluster-specific mean function and covariance function. The proposed algorithm comprises the probabilistic functional clustering (PFC) and the missing value imputation based on clustering results obtained from the PFC. The performance of the proposed method is demonstrated through a data example.