2026 Colloquia

Statistical Methods for Individual Precision in Functional Brain Imaging

Amanda Mejia, PhD
North Carolina State University

Functional magnetic resonance imaging (fMRI) enables noninvasive mapping of human brain function and is a valuable resource for neuroscience research and in clinical care. However, extracting accurate individual-level information from this high-dimensional, noisy, and biologically complex data remains a major challenge, which limits clinical translation. In this talk, I will present statistical methods developed to enhance individual precision in functional brain mapping by leveraging information from the population and across the brain. These methods are designed to respect key features of the data, including its spatiotemporal structure, neuroanatomy, and individual variability in both spatial and temporal brain organization. Finally, I will describe data science tools that facilitate adoption of these statistical techniques, including open source software, reproducible preprocessing pipelines, and resource sharing, with the ultimate goal of advancing the utility of functional neuroimaging for both research and clinical care.

Thursday, October 15, 2026
3:30 p.m. - 5:00 p.m.
Helen Wood Hall Room 1W-501

= = = = = = = = = = = = =

Beyond Euclidean Summaries: Distributional Methods for Wearable Device Data

Irina Gaynanova, PhD
University of Michigan

Wearable devices such as continuous glucose monitors and accelerometers record thousands of measurements per subject, yet analyses typically reduce these recordings to a handful of scalar summaries, discarding most of the available information. An appealing alternative is to treat each subject's measurements as a sample from a subject-specific distribution and to take that distribution itself as the unit of analysis. Doing so moves the problem out of Euclidean space: distributions equipped with a transport-based metric form a metric space rather than a vector space, so standard tools for regression, dimension reduction, and interpretation do not directly apply. In this talk, I will discuss statistical and computational challenges that arise in this setting, including efficient estimation and variable selection with distribution-valued responses, construction of interpretable low-dimensional summaries that respect the underlying geometry, and extensions beyond the univariate case where optimal transport becomes computationally prohibitive. The methods will be illustrated on data from continuous glucose monitoring studies.
Thursday, October 8, 2026

= = = = = = = = = = = = =

Bayesian Case Studies in Biomedicine

Michael Newton, PhD
University of Wisconsin-Madison

Visit the Andrei Yakovlev Colloquium for details.

Thursday, September 24, 2026

= = = = = = = = = = = = =

A Proxy Window into the Beautiful and Challenging World of Causal Inference

Eric Tchetgen Tchetgen, PhD
University of Pennsylvania

Visit the Charles L. Odoroff Memorial Lecture for details.

Thursday, May 7, 2026

= = = = = = = = = = = = =

Flexible Models for Multi-pollutant Mixtures and Health

Glen McGee, PhD
University of Waterloo

An important goal of environmental health research is to assess the risks posed by mixtures of multiple environmental exposures. In these mixtures analyses, flexible models like Bayesian kernel machine regression and multiple index models are appealing because they allow for arbitrary non-linear exposure-outcome relationships. However, this flexibility comes at the cost of low power, particularly when exposures are highly correlated and the health effects are weak. The first part of this talk will discuss a multivariate index modelling strategy that borrows strength across exposures and outcomes by exploiting similar mixture component weights and exposure-response relationships. We then extend the proposed approach to the multiple index model setting where the true index structure is unknown. The second part of the talk will focus on distributed lag non-linear models (DLNMs) for time series analyses of mortality and a mixture of multiple air pollutants, in which the timing of exposure effects is unknown. We propose a unified framework for multiple exposure DLNMs, integrating model specification, estimation, selection and stacking. The framework applies to four different model structures: two additive and two proposed single-index DLNMs for general outcome types. We develop a common estimation approach that applies to all four models, and we extend a model stacking approach to combine inferences across the four DLNMs.
Thursday, April 23, 2026

= = = = = = = = = = = = =

Model-assisted Reinforcement Learning

Eric Laber, PhD
Duke University

Reinforcement learning approaches are categorized as model-based or model-free by their dependence on a system dynamics model. We propose a model-assisted estimator that combines the advantages of model-based estimators with the model-free framework of V-learning. We show that this estimator is doubly-robust, being consistent when either (but not necessarily both) a model-free estimator or a model-based estimator is correctly specified. Furthermore, it is efficient when both estimators are correctly specified. The model-assisted estimator exhibits considerable improvement over model-based and model-free estimators in experiments.
Thursday, April 16, 2026

= = = = = = = = = = = = =

Monte Carlo Inference for Semiparametric Bayesian Regression

Daniel Kowal, PhD
Cornell University

Data transformations are essential for broad applicability of parametric regression models. However, for Bayesian analysis, joint inference of the transformation and model parameters typically involves restrictive parametric transformations or nonparametric representations that are computationally inefficient and cumbersome for implementation and theoretical analysis, which limits their usability in practice. We introduce a simple, general, and efficient strategy for joint posterior inference of an unknown transformation and all regression model parameters. The proposed approach delivers (1) joint posterior consistency under general conditions, including multiple model misspecifications, and (2) efficient Monte Carlo (not MCMC) sampling for the transformation and all parameters for important special cases. We illustrate the methodology for simulated and real data analysis with semiparametric Bayesian linear models, quantile regression, and Gaussian processes. These tools apply across a variety of data domains, including real-valued, positive, compactly-supported, and discrete (e.g., count) data.
Thursday, March 12, 2026