Teaching
Courses, invited talks, and openly available materials on statistical methods for high-dimensional biomedical data.
Courses
Taught in the Department of Epidemiology and Biostatistics.
EPIB 674
Statistical Foundations of Machine Learning in Public Health
Spring 2025, Spring 2026
The statistical reasoning underneath modern machine learning, taught for public health research: what these models estimate, when they generalize, and how to tell a real signal from an artifact of a high-dimensional dataset.
Currently teaching
EPIB 315
Biostatistics for Public Health Practice
Fall 2024, Fall 2025
Core biostatistical methods for public health practice: study design, estimation, inference, and the interpretation of published quantitative evidence.
Currently teaching
Invited talks & workshops
Recent invited presentations at statistical and biometric meetings.
-
Predicting Mortality in Older Men Through Multi-Omics Integration
Jun 2026
Multi-omics
-
Predicting Mortality in Older Men Through Multi-Omics Integration
Mar 2026
Multi-omics
-
Microbiome Mediation Analysis Through Absolute Abundances
Aug 2025
Causal inference
-
MetVAE: A Deep Learning Framework for Confounding Correction and Molecular Co-Occurrence Inference in Metabolomics Data
Dec 2024
MetabolomicsDeep learning
-
Multigroup Analysis of Compositions of Microbiomes with Covariate Adjustments and Repeated Measures
Jul 2024
ANCOM-BC2
-
Differential Abundance and Correlation Analysis of Microbiome Data: Challenges and Some Solutions
Apr 2021 · Workshop
Workshop
Learning to use our methods
The most-used teaching resource from this lab is not a course: it is the package documentation. The ANCOMBC vignettes walk through a complete differential abundance analysis on real data, from a phyloseq or TreeSummarizedExperiment object to interpreted results, and are written to be followed by someone with one semester of statistics.
MetVAE is documented the same way, with worked examples that take an untargeted metabolomics dataset through confounding correction to a correlation-based molecular network.
ANCOMBC vignettes MetVAE on PyPI All software
Teaching philosophy
Methods courses fail when students can derive an estimator but cannot tell whether it applies to the dataset in front of them. Our teaching pairs the mathematics with real, messy data and working code; students leave able to justify a modeling choice to a collaborator, not only to a grader.