The purpose of this project is to continue the development of new statistical tools and methods that address real issues of concern in empirical social science research. The primary focus of the prior work was on multiple imputation technology, which includes a family of techniques dealing with incomplete data. These techniques have been used increasingly by government agencies such as the Census Bureau and the Internal Revenue Service. Applications also have grown in academic research on a wide range of topics. Incompleteness, in fact, pervades intensively used scientific data as far ranging as social surveys and clinical medical protocols. Four main areas are addressed in the proposed research. First, continuing efforts will be made to hone the tools for multiple imputation. Second, modern methods of statistical computation such as extensions of the EM algorithm and the Gibbs sampler, will continue to be pursued. Third, the foundations of causal inference and practical methods for drawing causal inferences in observational studies will be further developed. And fourth, a variety of other techniques relevant to social science research, such as meta-analysis and mixture modelling, will be explored. The principal investigator's statistical contributions in the broad areas covered by this project have been impressive both in depth and volume. In addition, the many students he has trained in this long-term research agenda are superbly equipped to disseminate the results and to join in advancing the state of the art. The prospect of his achieving further significant accomplishments is very high.