David Kepplinger

David Kepplinger is an Assistant Professor in the Department of Statistics at George Mason University's School of Computing. His academic expertise spans robust statistical methods for high-dimensional data, computational statistics, and applications in biomedical sciences and environmental modeling.

Education:

• PhD in Statistics, University of British Columbia, 2020
• Master of Science in Statistics, Vienna University of Technology (Austria)

Dr. Kepplinger's research primarily focuses on robust estimation in high-dimensional settings, with particular interest in the robustness of feature selection in the presence of arbitrary contamination and countering the effects of contamination on predictive models. His work addresses critical challenges in statistical analysis where outliers and unusual values can severely impact results, especially in settings with many variables. He develops methods that maintain reliable performance even when data contains contamination in both response variables and explanatory features.

His recent publications demonstrate a strong interdisciplinary approach, spanning computational statistics, biomedical applications, environmental modeling, and clinical research. The research shows a consistent theme of developing robust statistical methodologies applicable to complex real-world problems across multiple domains, from protein biomarker identification to phenological modeling of cherry blossom predictions. His work frequently involves developing computational algorithms that can handle non-convex optimization problems common in robust statistics.

Dr. Kepplinger serves as a co-organizer for the statistics seminar series and is the PR & communications contact for the Department of Statistics at George Mason University. He mentors PhD students Yang Long and Siqi Wei, who focus on improving computational methods for robust regularized regression estimators. He teaches several courses including STAT 665 (Categorical Data Analysis) and STAT 778 (Statistical Computing).

He leads several notable projects including the First International Cherry Blossom Prediction Competition, which has been featured in prominent media outlets like The Weather Network, CBC Radio, and Public Radio's The World. This citizen science initiative aims to improve phenological modeling through crowd-sourced predictions of cherry blossom peak bloom dates in Washington, D.C., Vancouver, Kyoto, and Liestal-Weideli, Switzerland. He also develops multiple R packages available on CRAN and Bioconductor, including pense for robust regression estimation and examinr for creating online exams from R markdown documents.