EXPLOITING HIDDEN STRUCTURES IN EPIDEMIOLOGICAL DATA

点击翻译按钮获取中文摘要

Improving the accuracy and scope of categorical regression modeling performance would be invaluable to alcohol-related research and the overall health care community. Epidemiological models that provide better statistical significance and predictive performance enhance the researcher's ability to identify new patterns of alcohol-related symptoms as well as improving analysis of medical and psychiatric conditions in existing databases. Theoretical research and empirical evidence based on categorical data analyses for alcohol-related databases shows that substantial improvement in modeling results can be achieved by optimizing the data representation (recoding) scheme for selected predictor (explanatory) variables in a model. Martingale Research will develop an advanced data recoding algorithm utilizing techniques combining pattern recognition, stochastic optimization, and genetic algorithms to exploit structural relationships between predictor (continuous, categorical) and categorical outcome variables in a principled manner. This study develops an automated recoding optimization algorithm and demonstrates the algorithm using a database representative of pre-existing NIAAA sponsored databases. The project will show that an advanced recoding algorithm improves reliability and validity for a large class of categorical data regression models. These results will provide the essential first steps for additional investigations of dataset recoding for Phase II and form the foundation for developing a commercially available data analysis software package. PROPOSED COMMERCIAL APPLICATION: Martingale Research Corporation intends to develop database recoding algorithms into a commercial software package. These algorithms are intended to improve the overall performance of categorical models designed to explain data frequently encountered in the health care field. This approach applies as well to other industries that utilize categorical modeling to do financial prediction, risk analysis, and information interpretation and management.