Measurement errors in occupational epidemiology.
Source
NIOSH 1995 Oct; :1-14
Abstract
The specific aim of this project was the development of new measurement error methods which would be applicable to retrospective cohort and cross sectional studies typically found in occupational epidemiology. The methods were to be easy to use and understand. The usefulness of the methods was demonstrated through an analysis of an important occupational data set, the ACE study of the relationship between health and occupational exposure to anticancer drugs. Work was also performed on developing a user friendly computer software system to implement these methods so as to encourage routine use by occupational epidemiologists. The major findings of the study were that exposure measurement error is often an important source of bias in occupational epidemiology and that methods are available to correct for these biases. The second significant finding was that the fully parametric maximum likelihood method is efficient and consistent if an empirically verified measurement error model is correctly specified. Semiparametric estimating equations are locally efficient and can be used to check for sensitivity of the results to measurement error model misspecification. Thirdly, it was determined that automatic differentiation is an easy to use indispensable tool for developing an error free code for finding the root of nonlinear equations, such as parametric estimating equations, or for finding the maximum of the log likelihood.
Keywords
NIOSH-Grant; Grants-other; Epidemiology; Statistical-analysis; Risk-factors; Occupational-exposure; Antineoplastic-agents; Mathematical-models
Contact
Harvard University, School of Public Health, Department of Epidemiology and Biostatistics, 677 Huntington Avenue, Boston, MA 02115
Document Type
Final Grant Report
NTIS Accession No.
PB97-105662
Identifying No.
Grant-Number-K01-OH-000106
Priority Area
Grants Other
Source Name
National Institute for Occupational Safety and Health
Performing Organization
Harvard University, School of Public Health, Department of Epidemiology and Biostatistics, Boston, Massachusetts
