This project will add to Stata a module for Grade of Membership analysis. Grade of Membership (GoM) is a dimension-reduction tool for uncovering latent structure implicit in categorical data and, as such, is of most interest when utilized with datasets containing many categorical variables, such as data from the National Long Term Care Survey. There is currently no implementation of GoM in any general purpose, commercial statistical software package. This has been a hindrance both to applied researchers wishing to fit GoM models and to researchers interested in GoM methodology itself. The goal of the Phase I project is to develop, test, and certify two prototype Stata programs which fit two varieties of the GoM model, the conditional GoM model and the unconditional GoM model. The prototypes will not be optimized for speed and efficiency, but will instead serve to work out the algorithmic details of parameter estimation for both models and to certify the correctness of obtained results. Ultimately, these prototypes will be converted into software suitable for commercial release, including full documentation.
