Modern studies of health no longer observe people once. Smartwatches record movement and heart rate every second, smartphones ask about mood and sleep several times a day, glucose monitors and EEG headbands stream physiology, and location tracking connects all of this to weather, light, and greenspace. The result is intensive, multilevel, multimodal, longitudinal data collected across space and time — and it rarely fits the assumptions of classical models.
The CADENCE Lab — Contextual Analytics of Digital & Environmental data for Neurobehavioral, Circadian & Emotional health — builds the statistical theory and methods needed to learn from these data. Our unifying framework is the multivariate stochastic process: we treat each person’s streams of measurements as coupled random processes over time (and space), and develop models that respect their mixed scales (continuous, truncated, ordinal, binary), their dependence structure, and the context in which they were measured. We then use these tools to understand how mental health, sleep, physical activity, circadian rhythms, and the environment interact and evolve, and to move toward personalized prediction and early intervention.
Multivariate stochastic processes in space and time; semiparametric Gaussian copulas for mixed-type data; multivariate functional PCA for continuous, truncated, ordinal and binary functional data; dynamic structural equation models; graphical models and graph-constrained analysis; scalable multivariate spatial processes (graphical Gaussian processes, bigraphical Matérn–Whittle processes).
Objective streams from wearables — physical activity ⌚, heart rate ❤️, headband EEG 🎧, continuous glucose monitoring 🩸 — combined with subjective ecological momentary assessment 📱 of mood, energy, sleep and stress. We develop digital phenotypes for mood disorders and methods for real-time, intensive longitudinal mHealth data.
Linking health dynamics to where people are: outdoor temperature, daylight, greenspace and the built environment 📍, through location tracking and spatial statistics. Applications range from mood and sleep in mood-disorder subtypes to youth physical activity across social and physical contexts.