This research project addresses a series of methodological and theoretical statistical problems in the analysis of categorical data using loglinear and logistic response models, which grow directly out of problems in the study of the American educational system, and in basic educational research. The project focuses on the adaptation and development of general results on loglinear models to handle a wide variety of data-analytic situations involving qualitative data. Problems that are considered include: (1) modelling age, period, and cohort effects in the analysis of discrete archival data; (2) analysis and design problems related to the use of longitudinal and cross-sectional qualitative data; (3) simultaneous systems of logit models: (4) small sample properties of chi-square statistics; (5) models for mixtures of qualitative and quantitative variables; (6) asymptotic properties of chi-square statistics in large sparse contingency tables; (7) selection procedures for hierarchical loglinear model fitting; (8) the application of general loglinear model results to new problems, such as paired comparison models allowing for ties and order effects; (9) analysis of qualitative data from complex sample surveys, with special attention to cluster sampling; and (10) graphical methods for the presentation and analysis of multivariate data. (Author/GK)