Joseph L. Schafer

Joseph L. Schafer

Assistant Professor of Statistics

Ph.D.: Harvard University, 1991

Dr. Schafer has developed techniques for analyzing incomplete data and incorporating missing-data uncertainty into statistical inference. These techniques include both asymptotic approximations and simulation via multiple imputation, in which missing data are replaced by multiple simulated values. He has worked to develop general-purpose algorithms and software for the analysis of incomplete multivariate data. He has also been extensively involved with several projects undertaken by the Bureau of the Census, including the Post-Enumeration Survey to measure the undercount in the 1990 census. His research interests currently include the formal assessment of uncertainty due to missing data and other sources of assessment of uncertainty due to missing data and other sources of nonsampling error in sample surveys, techniques for computation and simulation of Bayesian posterior distributions, and parametric inference in sample surveys.

Representative Publications:

Rubin, D. B., J. L. Schafer, and N. Schenker. 1988. Imputation strategies for estimating the undercount. Proceedings of the Fourth Annual Research Conference, Bureau of the Census 151-159.

Rubin, D. B., J. L. Schafer, and N. Schenker. 1988. Imputation strategies for missing values in postenumeration surveys. Survey Methodology 14:209-221.

Rubin, D. B., and J. L. Schafer. 1990. Efficiently creating multiple imputations for incomplete multivariate normal data. Proceedings of the Statistical Computing Section of the American Statistical Association.