Variance component score test for multivariate change point detection with applications to mobile health

2026-01-29

AI Summary

This study addresses the challenge of real-time change-point detection in multivariate time series under high-dimensional, small-sample settings. The authors propose a variance component score test (VC*) that estimates distributional parameters exclusively from pre-change observations to efficiently detect abrupt shifts in the mean and/or variance. By restricting parameter estimation to data preceding the potential change point, VC* substantially reduces estimation bias and significantly enhances detection power in high-dimensional, low-sample scenarios. Theoretical analysis and simulations demonstrate that VC* achieves superior statistical power compared to existing methods. Furthermore, its practical utility is validated through successful identification of critical change points in smartphone-based behavioral data from adolescents exhibiting emotional instability, underscoring its applicability in real-world settings.

Technology Category

• Data Mining & Knowledge Management: Anomaly/Outlier Detection
• Machine Learning: Time-Series/Data Streams
• Intelligent Robots: State Estimation

Application Category

• Security and Privacy: Large-scale security measurements
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods

Abstract

Multivariate change point detection is the process of identifying distributional shifts in time-ordered data across multiple features. This task is particularly challenging when the number of features is large relative to the number of observations. This problem is often present in mobile health, where behavioral changes in at-risk patients must be detected in real time in order to prompt timely interventions. We propose a variance component score test (VC*) for detecting changes in feature means and/or variances using only pre-change point data to estimate distributional parameters. Through simulation studies, we show that VC* has higher power than existing methods. Moreover, we demonstrate that reducing bias by using only pre-change point days to estimate parameters outweighs the increased estimator variances in most scenarios. Lastly, we apply VC* and competing methods to passively collected smartphone data in adolescents and young adults with affective instability.

Problem

• multivariate change point detection
• variance component
• mobile health
• high-dimensional data
• distributional shift

Innovation

• variance component score test
• multivariate change point detection
• pre-change point estimation
• mobile health
• high-dimensional time series