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

TL;DR

This paper introduces a variance component score test (VC*) for detecting distributional changes in multivariate time series, especially in mobile health data, demonstrating superior power over existing methods through simulations and real-world smartphone data applications.

Contribution

The paper presents a novel VC* test that uses only pre-change data for parameter estimation, improving change point detection in high-dimensional mobile health data.

Findings

VC* outperforms existing methods in power during simulations.

Using pre-change data reduces bias and improves detection accuracy.

Application to smartphone data demonstrates practical utility.

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.