Creating Commercial Parallel Statistical Software

点击翻译按钮获取中文摘要

DESCRIPTION (provided by applicant): This project will create for market a large version of the Stata Statistical Software system for use on Symmetric Multiprocessor (SMP) computers that will execute more quickly than any other version of Stata. This new version will be built on top of Stata/SE, the current top-of-the-line version of Stata Statistical Software. Many health researchers regularly perform computationally intensive statistical analyses. Although computers have become very fast, these researchers are still constrained by the time it takes to run their analyses. The ready accessibility of SMP computers has thus far not been of use to these researchers because there is no commercial off-the-shelf statistical software that will perform its computations in a parallel fashion. In a previous Phase I SBIR project, Stata Corporation demonstrated that its method of parallelizing statistical computations in loops over the data can decrease run times of some statistical analyses on large datasets by almost half on a two-processor computer and to nearly a quarter on a quad-processor computer. In this project, Stata Corporation proposes to implement this methodology throughout all the parts of Stata that will benefit.