9310101 Clogg A five year program of research will be undertaken by the investigator to study seven interrelated statistical models for analyzing categorical dependent variables in social research. The objectives of the research are to: (1) Apply Rasch type models of the latent class variety to panel data so that unobserved heterogeneity can be characterized without extra assumptions about the form of the left out factors creating unmodeled variation; (2) Further developments in latent class analysis and measurement models; (3) Create new models and methods for regression with an ordinal dependent variable; (4) Generate models and methods for regression with a partially ordered Y; (5) Explore a new index of model fit based on mixture methods, with applications to social research; (6) Improve prediction methods associated with latent class models and apply these new prediction methods to graphic models; and (7) Model "total uncertainty" in statistical models used in social research. %%% This research program will generate and improve tools for a more rigorous empirical analysis in social research and provide results which will be of general interest to those in statistics and other areas. ***
