QBS 126 Analysis of Densely Collected Longitudinal Data
Rapid advances in technology has increasingly allowed for the collection of dense longitudinal data (i.e. data collected using many repeated measurements), and this type of data now abounds within biomedical and social science research (e.g., heart rate sensors, accelerometers, electronic medical record patient visits). A large variety of tools have emerged to model and predict dynamics that evolve over the course of time. The current course will discuss tools focusing on (1) explainability and theory-testing of dynamic processes with applications towards causal inference (e.g. multilevel models, vector autoregressive models, frequency domain analysis, state-space models, person-specific data models, dynamical systems modeling, varying-coefficient models, continuous time models) and (2) maximizing predictive performance (e.g. unique considerations in cross-validation with time-series data, time-series feature engineering, nomothetic and person-specific machine learning models, recurrent neural networks). Given the breadth of the tools in this field, the focus of this course will be primarily applied. Students will need to utilize both R and Python for this course.
(1 unit)
Instructor
Dr. Nicholas Jacobson