|
Activity Number:
|
414
- Risk Modeling and Regression Techniques
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Thursday, August 12, 2021 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Biometrics Section
|
|
Abstract #318813
|
|
|
Title:
|
Identification of Subgroups Using Multivariate Longitudinal Data: A Comparison of Model-Based Clustering, K-Means Clustering, and Deep Learning Approaches
|
|
Author(s):
|
Benjamin E Leiby* and Jobayer Hossain
|
|
Companies:
|
Thomas Jefferson University and Nemours Children's Health System
|
|
Keywords:
|
Longitudinal;
Clustering;
Data science;
Deep learning;
Mixture models
|
|
Abstract:
|
Identification of subgroups of patients based on patterns of change in multiple longitudinally measured outcomes has attracted much interest in the statistical and data science literature. Methods such as growth curve mixture models (implemented in Mplus software and the R package lcmm) and extensions of k-means clustering(R package kml3d)have been proposed. Researchers have also recently proposed the use of neural networks for clustering longitudinal data, although application in the multivariate setting has been limited. Through data analysis of multiple clinical measures from a cohort study of Glaucoma patients, we compare the strengths and limitations of these various techniques and discuss possibilities for two-step approaches that combine capabilities of multiple approaches.
|
Authors who are presenting talks have a * after their name.
