Grade of Membership (GoM) is a multivariate analytic technique for analyzing high dimensional discrete response data with very general distributional assumptions. GoM has been developed as a research tool at Duke University over the last 10 years with efforts focused on developing and assessing the statistical foundations of GoM. Little effort has been directed to producing a fully documented version of GoM. Thus, various versions of GoM exist with different capabilities. The first project task will be to identify an appropriate research version of the GoM which will be enhanced and documented for distribution to NIA and NIH researchers. The GoM version will be capable of handling datasets with sizes on the order of 10,000 observations and 30 analytic variables and will minimally operate on workstations and 486-based PCs. Such a version of GoM would be capable of analyzing many of the national survey datasets including the National Long Term Care Surveys. An intelligent support environment for the GoM model is also proposed to assist researchers in applying GoM. The support environment will include three modules: Question and Answers (Q&A), Data Preparation, and Report Analysis. Development of a prototype environment and a DOS version of GoM is proposed for Phase I.
