Research

Research Interests

• Non-tabular knockoffs variable selection
• Sports analytics
• Signal and image processing

My primary research interests center around the methodology of knockoffs variable selection on non-tabular data, specifically involving images from many contexts. Under the guidance of Dr. Zhe Fei, my PhD work focused on vision-transformer masked-autoencoder (ViTMAE) models to generate latent representations of images that can be used in an FDR-controlled variable selection framework called knockoffs to identify binary outcomes. As an application, I explored how this framework can be used to identify patients who exhibit signs of glaucoma.

Current research centers on embedding the framework under mathematical guarantees and extending it to other disciplines, like those seen in the medical and environmental fields.

Publications

Bouris E., Levya B.K., Odugbo O.P., Lawson J., Jin S., Fei Z., Morales E., Alkhalili O., Caprioli J. (2026). A Vision Transformer Model for the Detection of Early Glaucoma from Optic Disc Photographs. Scientific Reports, 16, Article 14831. https://doi.org/10.1038/s41598-026-44662-7

Lawson J., Fei Z. (2025). ImgKnock: Novel Knockoff Inference for Image Data via Latent Representation Learning. UC Riverside. Retrieved from https://escholarship.org/uc/item/77403004.

Lawson J., Watkins J. (2020). Legitimate or Unfair?: An Evaluation and Improvement of the College Football Playoff Using Logistic Regression and Adjacency Matrices. University of Arizona Campus Repository. Retrieved from: https://repository.arizona.edu/handle/10150/651357.