Skip to main content
ARS Home » Research » Publications at this Location » Publication #195585
Title: REGRESSION MODELING AND BEYOND--CHARTERING THE COURSE FOR GOOD HEALTH

Author
item SIMPSON, PIPPA - ACHRI
item GOSSETT, JEFF - ACHRI
item HUANG, BEVAN - ACHRI
item ROBBINS, JAMES - ACHRI
item CONNELL, CAROL - ACHRI
item CASEY, PATRICK - ACHRI
item Bogle, Margaret
item JO, CHANHEE - ACHRI

Submitted to: International Society for Behavioral Nutrition and Physical Activity
Publication Type: Abstract Only
Publication Acceptance Date: 4/20/2006
Publication Date: 7/13/2006
Citation: Simpson, P.M., Gossett, J.M., Huang, B.M., Robbins, J., Connell, C., Casey, P.H., Bogle, M.L., Jo, C. 2006. Regression modeling and beyond--chartering the course for good health [abstract]. Proceedings of International Society for Behavioral Nutrition and Physical Activity. p. 205.

Interpretive Summary:

Technical Abstract: Purpose: Nutrition data is measured with error. Yet typically in regression models, dietary information is included as an independent variable, assuming no measurement error. Structural equation modeling is a natural extension of regression modeling that allows incorporation of that error. Moreover, in the situation where diet may be intermediary between other variables such as demographics and obesity, more complex structures can be incorporated. Recently advances in software have allowed the consideration of complex weighted samples, including many national data sets. We show how interesting relationships between demographics, nutrition, health, and other personal characteristics may be investigated. Method: Using data from National Health And Nutrition Evaluation survey (NHANES) we illustrate the methodology using MPlus software. We show how to make explicit relationships between variables and diagram a model. We use a stepwise procedure to investigate parts of the model, using a traditional regression approach. We further show how different results may occur when measurement error is considered. In addition we show how separate estimates of the direct and indirect effects through nutrition, for example of age on obesity, can be estimated. Conclusion: Structural equation modeling allows better investigation of the interrelationship of nutrition, physical activity, health, and other personal characteristics.