Dynamic Functional Classification and Clustering

• Activity Number: 649
• Type: Contributed
• Date/Time: Thursday, August 4, 2016 : 8:30 AM to 10:20 AM
• Sponsor: Biometrics Section
• Author(s): J. Richard Landis and Wensheng Guo and Xiaoling Hou*
• Companies: University of Pennsylvania and University of Pennsylvania Perelman School of Medicine and University of Pennsylvania
• Keywords: Dynamic State Space model ; Functional Mixed Effect Models ; Smoothing Spline ; Urological chronic pelvic pain

Abstract

We propose a dynamic functional clustering algorithm where each group of curves are modeled by a functional mixed effect model and the posterior probability is used to iteratively classify each subject into different groups. The functional mixed effects model allows flexible designs and nested structures. The classification takes into account both group-average trajectories and between-subject variability. We propose an equivalent dynamic state space model to calculate the likelihood in fitting the model and to efficiently compute the posterior probability in classifying a new subject. The resultant sequential algorithm is $O(n)$ and can be implemented online. We also propose a leave-one-subject-out cross-validation Kullback-Leibler information criterion to choose the number of clusters. The performance is assessed through a simulation study and we apply the proposed methods to the urological chronic pelvic pain syndrome symptom (UCPPS) data collected from the study of Multidisciplinary Approach to the Study of Chronic Pelvic Pain (MAPP) research network and identify three sub-groups.