Dr. J. Sunil Rao recently received a methods research grant where he proposes to develop effective methods for data analysis and prediction in vital areas of application, from privacy protection via differential privacy (DP), to precision medicine and public health disparities. The areas of application in the grant will range from the prediction of epigenetic markers in cancer using genetic data, to predictions with employment data from the U.S. Bureau of Labor Statistics (BLS).

The study, titled “Modernizing Mixed Model Prediction," is funded by the Division of Mathematical Sciences at the National Science Foundation, and will run until 2025.

The goal of this project is to develop and apply new methods in order to expand the range of problems that can be handled by mixed model prediction (MMP), which is a technique traditionally used for correlated data settings. “Particularly, for the DP application, we will apply the methods to the publicly released 2020 U.S. decennial census” said Rao, Ph.D., Principal Investigator, professor, and director of the Division of Biostatistics at the University of Miami.

“For the BLS application we will target questions regarding volatility during the ongoing COVID-19 pandemic that thus require robust modifications from traditional approaches. This grant is a continuation of NSF funding for my work in MMP and reflects how important the funding agency views making advancements in these types of prediction problems,” added Dr. Rao. 

Investigators will concentrate on three major aims for this project.

• Multivariate mixed model prediction (MMP) in genomic prediction problems where correlated DNA methylation markers, which are one form of epigenetic markers, reflect underlying disease biology and improved prediction accuracy is possible by borrowing strength across this multivariate structure.
• MMP for differentially private (DP) data in which cluster or grouping identities are contaminated by design and not released to protect privacy. It is a technique that enables obtaining information about groupings within the dataset, yet keeps personal identification private.
• MMP with non-Gaussian, also known as non-normal distribution, random effects and errors, which can prominently expand the range of circumstances in which MMP can be applied beyond the classical normality assumptions that do not fit many modern datasets. “It extends the usual models to a broader range of settings by relaxing the usual assumptions (Gaussian random effects and errors). This is important for a number of modern datasets which are often more complex in nature where the older assumptions are less likely to hold,” he said.

Dr. Rao and colleagues at UC-Davis and Oregon Health and Science University will develop the methodology necessary for each aim, conduct theoretical research on the procedures, and carry out extensive empirical simulation studies to compare the new methods with other methods. Furthermore, he will collaborate closely with subject-matter experts on putting the methods developed in this project to use in answering practical questions. “The grant will also involve the development of user-friendly software which will be made available to the broader research community,” concluded Dr. Rao.

Written by Deycha Torres Hernández

Published on September 21, 2022