PAPER ABSTRACT

Conference on Simulation in the Medical Sciences

Proceedings of the 1997 Western MultiConference

ISBN# 1-56555-105-2

This paper is copyrighted by Simulation Councils, Inc.

For reprints, please contact The Society for Computer Simulation International at P.O. Box 17900, San Diego, CA 92177, USA or at .

Using Constrained Categorical Regression to

Identify Structural Relationships in Epidemiological Data

Richard M. Golden

RMG Consulting, 2000 Fresno Road, Plano, Texas 75074

golden@utdallas.edu

Steven S. Henley, Harvey L. Bodine, Robert L. Dawes

Martingale Research Corporation, 1485 Richardson Drive, Suite 110, Richardson TX 75080

mrc@metronet.com

T. Michael Kashner

UT Southwestern Medical Center at Dallas, 8267 Elmbrook, Suite 250, Dallas, TX 75247-9141

KEY WORDS

Statistical Analysis, Categorical, Epidemiological, Modeling, Regression, Neural Network

ABSTRACT

Improved methods for confirmation of structural relationships between demographic information sources and medical and psychiatric conditions are invaluable to local, county, state, and national groups and agencies. A neural network based nonlinear regression model known as the CCR (Constrained Categorical Regression) model is introduced that provides an explicit mechanism for representing structural relationships as logical rules (i.e., Boolean functions) whose respective contributions to predicting explicit outcome probabilities are weighted by the model's free parameters. The CCR modeling approach combines in a novel way the classical generalized logit modeling methodology with relatively new statistical hypothesis testing techniques designed for solving problems in econometric and artificial neural network modeling. In this paper, the relationship of the CCR model to classical methods of statistical inference is discussed and the results obtained from applying the CCR model to the analysis of a particular epidemiological data set are presented.

REFERENCES

1. Tabachnick, Barbara G. and S. Fidell, 1996. Using Multivariate Statistics, 3rd Edition, Harper Collins College Publishers, New York, New York.
2. Stopher, Peter R., Arnim H. Meyburg. 1979. Survey Sampling and Multivariate Analysis for Social Scientists and Engineers, D.C. Heath and Company, Lexington, Massachusetts.
3. Golden, R. M. 1995. Making correct statistical inferences using a wrong probability model. Journal of Mathematical Psychology 38, 3-20.
4. Golden, R. M. /in press/. Mathematical methods for neural network analysis and design, MIT Press, Cambridge, MA.
5. White, H. 1982. Maximum likelihood estimation of misspecified models, Econometrica 50: 1-25.
6. Dawes, R., S. Henley, H. Bodine, R. Golden, and T. Kashner. 1995. "Confirming and Exploiting Logical Causal Relationships between Alcohol-related Categorical Variables", Technical Report N43AA42009 for National Institute on Alcohol Abuse and Alcoholism.
7. Golden, R., S. Henley, H. Bodine, R. Dawes, T. Kashner, /in press/. "Pruning a Softmax Neural Network using Principled Optimal Brain Damage", In Proceedings of the 1996 World Congress on Neural Networks.
8. Bodine, H., S. Henley, R. Dawes, G. Golden, T. Kashner, /in press/. "Data Modeling using Constrained Categorical Regression", In Proceedings of the 1996 Artificial Neural Networks in Engineering.

NOTE: This material was based on work sponsored by the National Institute on Alcohol Abuse and Alcoholism. The opinions, findings, and conclusions or recommendations expressed in this publication are those of the authors and do not necessarily reflect the views of the National Institute on Alcohol Abuse and Alcoholism.