Research Grants/Fellowships/SBIR

Analysis of Nondifferential Exposure Misclassification in Ecological Epidemiology Using a General Error Model

EPA Grant Number: U914942
Title: Analysis of Nondifferential Exposure Misclassification in Ecological Epidemiology Using a General Error Model
Investigators: Webster, Thomas F.
Institution: Boston University
EPA Project Officer: Manty, Dale
Project Period: January 1, 1996 through January 1, 1999
Project Amount: $102,000
RFA: STAR Graduate Fellowships (1996) RFA Text |  Recipients Lists
Research Category: Fellowship - Probability/Statistics , Academic Fellowships , Environmental Statistics

Objective:

The objective of this research project is to develop statistical methods in environmental epidemiology.

Approach:

I analyzed the effects of nondifferential exposure misclassification in linear ecological regression using a general error model (Wacholder, Epidemiology 1995;6:157-161). The results of Brenner, et al. (NDEM of binary exposure) follow naturally from this perspective. Other important cases of the general error model arise as well. Use of the average exposure within a group is a form of Berkson error, causing no bias when other sources of ecological bias are absent. When external group measures of exposure are employed, (e.g., concentrations of pollutants in air or water), we can examine the difference between this group measure and the average exposure. The well-known classical error model (bias toward the null) arises when the difference and the average exposure are uncorrelated. In sum, the type of bias associated with ecological studies depends on the error structure.

Supplemental Keywords:

fellowship, statistical methods, environmental epidemiology, general error model, linear ecological regression, exposure, ecological bias, classical error model., RFA, Economic, Social, & Behavioral Science Research Program, Scientific Discipline, Applied Math & Statistics, Epidemiology, Environmental Statistics, nondifferential exposure missclasification, ecological epidemiology, statistical models, regression analysis, error measurement techniques, general error model, data analysis