Paul Bailey

Paul Bailey specializes in studies with merged data, statistics, psychometrics, and educator value-added modeling. He is currently the head statistician for EdSurvey, an R statistical package AIR created to serve as a one-stop shop for downloading, processing, manipulating, and analyzing survey data from the National Center for Education Statistics (NCES). He has embedded use of plausible values, jackknife replicate weights, and Taylor series approximations into the EdSurvey package, substantially reducing the difficulty of using NCES data. He has also developed novel software for the estimation of weighted mixed models in the WeMix R package and for calculation of weighted correlations in another R package (wCorr). Additionally, he further developed a latent regression R package (Dire) that handles multidimensional constructs often used by large-scale assessments, ultimately allowing researchers to use published item parameters to estimate their own conditioning models and draw their own plausible values. This software is also useful for merged data, such as large-scale assessments when linked to administrative data.

Dr. Bailey recently completed work for Arnold Ventures on the return to education for veterans using GI Bill funding. This involved bringing together data from the Defense Manpower Data Center, Veterans Benefit Administration, Internal Revenue Service, U.S. Census Bureau, and the National Student Clearinghouse. This work culminated in several reports on veterans' uptake of the Post-9/11 GI Bill, college graduation, and labor market outcomes.

Previously, Dr. Bailey was AIR’s head statistician for New York State’s value-added model for educator evaluation, the results of which were used in annual performance reviews for principals and teachers throughout the state. In this role, he worked with the state to develop, evaluate, and maintain models and managed AIR’s computing staff that produced results used by the state.

Dr. Bailey has also performed psychometric analysis of student surveys and written psychometric technical reports, identifying constructs from previously unanalyzed surveys and characterizing evidence for the psychometric validity of existing constructs on other surveys.

Before becoming an economist, he worked for 7 years as a radiation physicist for the U.S. government, specializing in the measurement of ionizing radiation.