This Project Grant from the National Science Foundation's National Center for Science and Engineering Statistics will fund the development of Bayesian statistical and machine learning methodologies tailored for complex survey and census data. Awarded $743,050 under the Social, Behavioral, and Economic Sciences program, the grant will support research at the University of Missouri from September 2022 through August 2025.
The research aims to advance computational efficiency and expand analytical capabilities for important federal data sources like the Survey of Graduate Students and Postdoctorates in Science and Engineering. Methods will leverage randomization techniques within Bayesian hierarchical models (Aim 1) and apply random weight neural networks for nonlinear regression and data integration (Aim 2). Recurrent neural networks will also be used to model temporally correlated complex survey data (Aim 3). Successful development of these principled methodologies will benefit both scientific and federal statistical analysis of complex data in fields like demography, econometrics, and political science.