Statistical methods for the analysis of complex-structured biomedical data

We develop innovative statistical methodology that turns the structure in data into signal—explicitly leveraging complexities of data from modern biomedical technologies.

• Hierarchical Bayesian modeling to encode domain knowledge and to propagate uncertainty.
• Variable selection to pinpoint a parsimonious set of biomarkers from high‑dimensional feature sets.
• Geometric & functional data analysis to analyze shapes, curves, and trajectories.
• Spatial statistics to exploit spatial dependence in data.
• Network models in statistics & machine learning to learn interaction structure and to infuse graph constraints.

Together, these methods provide robust and interpretable inference and prediction—bridging complex structure to clinically meaningful insight while remaining firmly grounded in rigorous statistical principles.