Stat Cafe - Rebecca Lee
Hierarchical Bayesian Spectral Estimation of Stationary Time-Series of Varying Lengths
- Time: Monday 4/14/2025 from 11:30 AM to 12:30 PM
- Location: BLOC 457
Description
Biomedical time series data, such as heart rate variability, electroencephalography, and functional magnetic resonance imaging, provide important indirect measurements of physiological processes underlying health. Biomedical studies typically collect time series data from multiple participants to characterize the relationship among underlying biological processes, treatments, and health outcomes of interest. Although time series from different patients often exhibit similar dynamics, there is still considerable variability between patients due to individual differences. Bayesian hierarchical models allow us to bridge this divide by sharing information across participants exhibiting similar behavior while still allowing for potential variability in behavior across participants. We present a non-parametric frequency-domain hierarchical Bayesian method that allows for “borrowing” of information across replicates and calculates estimates for both the replicate-specific log-SDF and the global log-SDF for multiple stationary time series of varying lengths.
Gallery (with photos by Samantha Williams)
