Estimation of Radiation Dose from Biological Manifestations and Imperfect Measures of Physical Determinants

• Activity Number: 492
• Type: Invited
• Date/Time: Thursday, August 7, 2008 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Randolph L. Carter*+ and Harry Cullings and John Cologne and Sachiyo Funamoto and Yoichiro Kusunoki and Kazuo Neriishi and Thomas Seed and Austin Miller and Carmen D. Tekwe and Nori Nakamura and Daniel Stram and Norman P. Ross
• Companies: University at Buffalo and Radiation Effects Research Foundation and Radiation Effects Research Foundation and Radiation Effects Research Foundation and Radiation Effects Research Foundation and Radiation Effects Research Foundation and Radiation Effects Research Foundation and University at Buffalo and University at Buffalo and Radiation Effects Research Foundation and University of Southern California and Radiation Effects Research Foundation
• Keywords: measurement error ; Berkson error ; instrumental variable estimation ; estimation bias ; biodosimetry

Abstract

Current dosimetry system (DS02) estimates of atomic bomb radiation dose are based on physical determinants. Two distinct types of errors exist in DS02 values, with statistical characteristics of measurement error and Berkson error. With measurement error alone in a linear model, dose effects are not identified without additional information. We use biological manifestations of dose as instrumental variables to identify parameters. With Berkson error alone, the model is identified. With both types of error present, a known/estimated measurement error variance suffices to identify the model, but an instrumental variable alone does not. An instrumental variable plus estimated Berkson error variance suffices. We propose methods for estimating model parameters and for combining information in biological dosimeters with that in physical dosimeters to improve estimates of radiation dose.