Alcohol-Related Categorical Variables

Improved methods for identification of new longitudinal patterns of alcohol-related symptoms,administrative strategies, as well as medical and psychiatric conditions which effectively predict patientoutcome status in existing databases would be invaluable to local, county, state, and national groupsand agencies. We will develop constrained Categorical regression (CCR) software for integratingestablished principles of log-linear modeling, structured equation modeling, and rule-based systems intoa single software package. The CCR software improves the power of a statistical analysis by directlyincorporating the user's "intuitions" in a principled manner. These intuitions are formally representedas logical causal relationships. The proposed Phase I study will demonstrate using a databaserepresentative of pre-existing NIAAA sponsored databases: (1) that the CCR model is sensitive enoughto detect differences between distinct sets of logical causal relationships and patient/providersubpopulations, and (2) makes accurate predictions about the likelihoods of particular patient outcomes.The results obtained in the Phase I study provide the essential first steps for additional investigationsof model reliability, validity, and repeated measures for Phase II and form the foundation for developinga commercially available CCR software package.