Predicting Changes in Individual Wellbeing Scores: Mixed Effects Models using Sleep Data from Wearables

Sleep plays a major role in regulating human cognitive function, performance, mood, and well-being. Despite its significance, the intricate relationship between various sleep components—such as duration, quality, and regularity—and wellbeing outcomes remains inadequately explored. The nature of sleep data poses challenges in capturing and interpreting temporal patterns, but the growing popularity of wearable devices capable of collecting vast multi-modal data presents a promising avenue to bridge this gap. In this thesis, the aim is two-fold: first, identify the impact of different combinations and transformations of sleep regularity (Sleep Regularity Index- SRI, Composite Phase Deviation- CPD, Interdaily Stability- IS) and duration calculated from wearable devices across varying time frames on self-reported morning wellbeing scores (alertness, happiness, energy, health, calmness); and second, evaluate both linear and nonlinear associations between different sleep metrics and wellbeing. To address high user variability found by the personalized nature of sleep and the subjective nature of wellbeing assessments, we employ mixed effects modeling techniques where each individual is treated as their own cluster, including Linear Mixed Effects models (LMM) and Mixed Effects Random Forest (MERF), where the latter is benchmarked against classic machine learning models. The LMM results were most statistically significant for independent regularity (SRI, IS), combined regularity (SRI and IS), total sleep time as duration (TST), and combined regularity and total sleep time (SRI and TST, IS and TST) for alertness and energy over 2-4 nights. MERF outperformed other models in Mean Absolute Error (MAE), for all time split scenarios. This research further emphasizes the importance of addressing data leakage due to the time sensitivity of sleep data and calculation of regularity spanning multiple days. Bye stablishing correlations between sleep parameters and wellbeing indicators, this study hopes to provide deeper insights into fluctuations in wellbeing and inform the development of wearables that monitor sleep patterns.

Name thesis:degree_name

Master

Department dc:contributor.department

Massachusetts Institute of Technology. Department of Electrical Engineering and Computer Science

Grantor dc:publisher

Massachusetts Institute of Technology

Year dc:date.issued

2024

Author dc:creator

• Choi, Shelley

Advisor dc:contributor.advisor

• Picard, Rosalind

Statement dc:rights

• Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0)
• Copyright retained by author(s)

Licence dc:rights.uri

• https://creativecommons.org/licenses/by-nc-nd/4.0/

Handle dc:identifier.uri

https://hdl.handle.net/1721.1/156760

OAI identifier oai:identifier

oai:dspace.mit.edu:1721.1/156760