USU Doctoral Student Receives Top Recognition From American Statistical Association

Utah State University doctoral student Maha Moussa, a dual statistics and data science major, is a top awardee in an international statistics and data science student paper competition. As part of the recognition, she has been invited to present her research at North America’s largest gathering of statisticians and data scientists in summer 2026.

Moussa’s paper was among three selected for top recognition in the Section on Statistics and the Environment Student Paper Competition of the American Statistical Association. She and fellow awardees will share their papers at Joint Statistical Meetings 2026, Aug. 1-6, in Boston. The gathering, which is expected to draw more than 5,000 attendees from 52 countries, encompasses members of the ASA, as well as professional statistics and data science societies throughout the world.

“This is a highly competitive and prestigious honor,” says Yan Sun, Moussa’s mentor and associate professor and assistant department head of graduate studies in USU’s Department of Mathematics and Statistics. “Maha is one of the hardest working students I have ever met and is genuinely deserving of this recognition. She is on her way to an outstanding scholastic future.”

Moussa’s submitted manuscript, under peer review, details her development of a novel, AI-guided statistical inference framework that uses deep neural networks to enhance modeling and forecasting of large-scale spatiotemporal precipitation data. Co-authors on the paper are Sun, along with Wei Zhang, assistant professor in USU’s Department of Plants, Soils and Climate, and Shandian Zhe of the School of Computing at the University of Utah.

“Our paper goes beyond prediction,” Moussa says. “In addition to improving accuracy, it helps us understand which factors affect rain-snow transitions. Traditional models have limitations at this scale, so we combined classical statistical modeling with AI to benefit from AI’s high accuracy while maintaining the interpretability of statistical models.”

While classical varying-coefficient (VC) models are widely used in many areas of statistical analysis, she says, “they can struggle with the scale and dimensionality of large datasets spanning diverse landscapes.”

“We developed an AI-guided, single-index VC logistic framework, which delivers highly accurate results,” Moussa says. “This model enables analysis of rain-snow transitions at the continental scale.”

Accurately distinguishing rain from snow is fundamental for hydrologic prediction, flood and runoff forecasting, water-supply planning and climate diagnostics, she says.

“The model recovers spatially varying, physically ordered rain-snow temperature thresholds, learns a one-dimensional contextual index that organizes those thresholds across space, and retains state-of-the-art predictive skill while remaining interpretable,” Moussa says. “It can be used to classify rain versus snow, and also to uncover drivers of snow formation, including variations in elevation, temperature and humidity.”

Sun adds Moussa’s work demonstrates how AI and statistics can strengthen one another in future data science research.

“People may think traditional statistics is being replaced by AI, but that’s not the case,” she says.

Moussa agrees.

“Our model shows that deep learning and classical statistical modeling are complementary,” she says. “We are not replacing statistics — we rely on it to help open the ‘black box’ of AI.”