The last two decades have seen an explosion in computational capacity, and a corresponding reorientation of statistics and econometrics toward making intensive use of computation to reveal more about data and sharpen the inferences drawn. Statistical examples are the development of robust estimation methods, graphical interfaces for data examination, and bootstrap and other resampling techniques to sharpen finite-sample inferences. Econometrics examples are the development of semiparametric methods, nonlinear time series techniques, and simulation methods for inference. Incorporation of these methods into econometrics courses and applied econometrics software has been uneven, particularly for methods developed primarily in statistics. This grant supports research and teaching conferences on statistical computation and econometric applications in order to address this problem. The activities planned under this project promote the use of current statistical methods and UNIX workstation software in econometrics, encourage the augmentation of statistical software systems (such as the S language) with the procedures and modules needed for many standard econometric procedures, familiarize statisticians with some of the methodological and computational problems encouraged in econometrics applications, and encourage joint developments of computational methods. Three research conferences are planned, one in each year of the proposal. The conferences deal with current issues in computer- intensive methods in statistics and econometrics. Three teaching conferences will be held, again one in each year. These conferences are designed to provide an intensive introduction in methods and issues in statistical computation to groups of graduate students nominated by Departments of Economics and Statistics around the country. These activities are coordinated to achieve the objectives of (1) providing a statistical computing environment that integrates statistical and econometric software, with documentation and tutorials that encourage more effective use of existing software by econometrics and statistics faculty and graduate students, (2) identifying productive areas for research and development on statistical algorithms, and suggesting standards for econometric software, (3) encouraging the implementation of new algorithms in "open look" software, and (4) providing "how to" materials to the econometric research community on setting up and operating a statistical computing environment for research and teaching, and on econometric software standards.
