Proposal: DMS 9504589 PI: Suojin Wang Institution: Texax A&M Title: SOME ESTIMATION AND COMPUTATION PROCEDURES IN STATISTICS Abstract: The primary objective of this research is to develop new theory and methods in some important classes of statistical problems in the areas of sampling, missing data analysis and measurement error models. The research topics are motivated by practical problems and the main focus is on estimation and (approximate) computation procedures. The research aims to significantly advance statistical knowledge in these areas, and improve some current practices in survey agencies in particular. In attacking some of the proposed problems, small sample asymptotic methods are desired. The potential use of accurate saddlepoint approximations in these problems is studied. In other proposed problems, large sample asymptotic methods are investigated. The research topics include kernel-type estimations in regression when data are partially missing or measured with error; sub-domain estimation in sample surveys; and new and optimal estimation of the distribution function of a finite population. In this research project new statistical theory and methods are developed and investigated in statistical problems that bear practical importance. The accuracy and simplicity are the main criteria in developing such statistical methods. They have direct applications in various agencies, such as the Bureau of Labor Statistics, to improve the accuracy and speed of the current statistical practice. Furthermore, this research advances the science and technology in the Federal Strategic Areas of high performance computing and the statistical aspects of biotechnology.
