DOI

https://doi.org/10.25772/3ddf-c979

Author ORCID Identifier

https://orcid.org/0009-0003-6660-0391

Defense Date

2026

Document Type

Dissertation

Degree Name

Doctor of Philosophy

Department

Biostatistics

First Advisor

Shanshan Chen

Second Advisor

Nitai Mukhopadhyay

Third Advisor

Dipankar Bandyopadhyay

Fourth Advisor

Brian Berman

Fifth Advisor

Maha Alattar

Abstract

Actigraphy and polysomnography (PSG) represent complementary but largely disconnected modalities for measuring sleep. Actigraphy enables intensive longitudinal data collection at scale but lacks physiological detail, while PSG provides gold-standard sleep architecture characterization at the cost of logistical feasibility in large studies. This dissertation develops a unified statistical pipeline for extracting and analyzing night activity segments from actigraphy data and establishing their quantitative connections to PSG-derived sleep metrics. First, we model actigraphy under the Zero-Augmented Gamma distribution to account for its zero-inflated, right-skewed structure. Second, we introduce Pi-Change, a prior-informed multiple change point detection algorithm that incorporates circadian rhythm information via a time-varying penalty. Third, we develop a hypothesis test framework for detecting segmentation errors in actigraphy data. Fourth, we propose a novel marked point process model for night activity segments that accommodates inter-device variation in actigraphy. Finally, we apply a pipeline composed of the previous methods to the Multi-Ethnic Study of Atherosclerosis Sleep dataset, using multi-state models and Bayesian mixed-effects models to characterize associations between actigraphy-derived parameters and PSG sleep stage metrics, demographic covariates, and subjective sleep quality. Results indicate that the point process parameters representing nocturnal activity onset and duration are associated with sleep onset latency, time spent in the wake after sleep onset stage of sleep, total sleep time, and sleep efficiency. Further, these parameters are associated with age, sex, race/ethnicity, and job schedule. Together, these contributions provide a principled, scalable framework for bridging wearable and gold-standard sleep measurement, with implications for population-scale sleep research and clinical monitoring.

Rights

© The Author

Is Part Of

VCU University Archives

Is Part Of

VCU Theses and Dissertations

Date of Submission

8-7-2026

Available for download on Wednesday, August 06, 2031

Included in

Biostatistics Commons

Share

COinS