Ambulatory Big Data Methodologies

Is the mundane significant? We seek to understand how subtle processes are associated with important outcomes. Small, daily behaviors can have large impacts on functioning and are important points of intervention for improving child and family mental health wellness. In addition to collecting laboratory-based data, we collect data in families’ day-to-day lives to understand how family dynamics naturalistically unfold. Using smartphones and wearable devices, we generate ambulatory big data to track functioning in real time and real life.

Intensive Longitudinal Data Analysis

To study family dynamics, we collect intensive multimodal data and use novel computational methods for measuring, quantifying, and understanding interpersonal relationships. People are connected to each other through intricate and fast-moving patterns of behavior where information is communicated between people through a variety of modalities, including each person’s emotional state, language, tone of voice, physiological arousal, and body language. Longitudinal, cross-person, multimodal data such as these are highly complex. We apply state-of-the-art statistical models and develop new techniques and methodologies to quantify these patterns and processes.

Machine Learning and Mental Health Wellness

With the advent of wearable sensors and mobile technologies, vast amounts of data can now be collected relatively easily and inexpensively. However, methods for making sense of these vast amounts of data have lagged behind. Big data and machine learning are increasingly being used to learn patterns in such data to better understand psychological functioning and improve well-being. We apply and develop machine learning methods to automatically and passively detect and even predict interpersonal processes via wearable devices as they occur in daily life.

Intensive Longitudinal and Multimodal Data Analysis

Human behavior unfolds through complex processes that operate across multiple timescales, from moment-to-moment fluctuations to long-term developmental change. Traditional research methods often capture only a small fraction of this complexity. Advances in intensive longitudinal assessment and multimodal measurement create new opportunities to investigate how psychological, behavioral, interpersonal, and physiological processes evolve in everyday life.

Our lab applies and develops advanced statistical methods for analyzing intensive longitudinal and multimodal data, including repeated self-reports, behavioral observations, physiological signals, and naturalistic recordings. We are particularly interested in understanding how processes operating at different timescales interact, how patterns vary across individuals and contexts, and how small changes in everyday experiences contribute to longer-term developmental and clinical outcomes.

Dynamic Systems and Computational Modeling

Psychological functioning emerges from complex, interconnected systems in which emotions, behaviors, relationships, and environments continuously influence one another. These systems are characterized by feedback loops, nonlinear dynamics, and patterns of stability and change that cannot always be understood through traditional statistical approaches.

Our research draws on dynamical systems theory and computational modeling to investigate how psychological and interpersonal patterns emerge, persist, and transform over time. We use approaches such as time-series analysis, multilevel modeling, state-based models, and computational methods for characterizing transitions between patterns of functioning. By identifying the processes that maintain stability or facilitate change, we seek to advance theoretical understanding of psychological development, psychopathology, and resilience.

Machine Learning and Predictive Analytics

The growing availability of complex behavioral and physiological data creates opportunities to identify patterns that may be difficult to detect using conventional analytic approaches. Machine learning and predictive modeling offer powerful tools for integrating information across multiple sources, identifying meaningful differences between individuals, and anticipating changes in psychological functioning.

Our lab applies machine learning and computational approaches to characterize behavioral and physiological patterns, identify latent states of functioning, and predict clinically meaningful outcomes. We are particularly interested in combining data-driven discovery with psychological theory to develop interpretable models that reveal mechanisms of risk, resilience, and change. Ultimately, our goal is to advance computational methods that improve scientific understanding, generate new hypotheses, and support more precise approaches to mental health research and intervention.