STATISTICAL ANALYSIS OF BODY COMPOSITION RISK FACTORS

描述：（改编自申请人描述）：建议

DESCRIPTION: (Adapted from Applicant's Description): This proposed research is a continuation and expansion of preceding work performed by this applicant in a current project. The overall goals are to apply recent, advanced, statistical methods to studies of human body composition and risk factors for cardiovascular and related diseases. Existing data from a longitudinal study of cardiovascular disease risk factors, including body composition measurements, will be used to test the applicability of the advanced statistical methods. Two Specific Aims of this application are: 1) to determine if statistical models or adjustments can be used to account for the effects of measurement errors in longitudinal studies, regarding the regression estimates, hypothesis testing, and model predictions; measurement errors can reduce relationships between dependent and independent variables resulting in biases for regression estimates, reduced validity of hypothesis testing, and reduced generalizability of the prediction of future risk; and 2) to evaluate the simultaneous use of several risk factors in multivariate, longitudinal models to improve the prediction of future risk. These proposed statistical refinements are motivated by the study of long-term serial risk factors for cardiovascular diseases and the problems generated by measurement errors introduced into these studies, in particular those errors associated with body composition measurements. The impact of measurement errors in independent variables on longitudinal models will be examined and adjusted. The ability to accurately predict future values from earlier values, such as in childhood, is important in many longitudinal studies, and may help to identify or predict those values associated with increased risk. Nevertheless, patterns of change in one variable can be influenced by other variables, hence statistical modeling of the multivariate relationships among risk factors should aid in the derivation and clarification of relationships among multiple variables, and allow for improved prediction models. The investigator's application is that the simultaneous use of several risk factors (c.f., a single risk factor) will improve the identification of individuals at future risk. The model development proposed in this work will use existing data for body composition and risk factors for cardiovascular and related diseases from the Fels Longitudinal Study. The application points out that the models to be developed are not limited to these data, but can be used in the analyses and interpretation of other longitudinal data sets.