Director, Innovation Policy RTI International Washington, District of Columbia, United States
Dr. Alexander will present current work at RTI International to develop and deploy an new approach to inferring the demographic characteristics of research funding applicants and awardees through Bayesian reasoning and data linkage. At agencies such as the U.S. National Science Foundation, efforts to identify individuals from underrepresented groups in research funding are stymied by missing data, especially in cases where applicants and peer reviewers decline to share information. To address the missing information that is needed to assess efforts targeting the "missing millions," RTI is combining administrative data from proposal and grant management systems with other statistical databases (including the Integrated Postsecondary Education Data System and the Doctoral Records File) to infer characteristics of individuals from variables such as geographic location, institutional affiliation, and educational background. This talk will describe the Bayesian Improved Surname Geocoding process to impute missing race/ethnicity data using multiple data sources, and review the opportunities and challenges in using this approach to inform evaluations of efforts to broaden participation in scientific research.
