|
Activity Number:
|
213
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Tuesday, August 5, 2003 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Statistics in Epidemiology
|
| Abstract - #302205 |
|
Title:
|
Weighting Methods to Adjust for Nonresponse to Epidemiologic Surveys in the Presence of Unknown Eligibility
|
|
Author(s):
|
Timothy R. Church*+ and Steven J. Mongin
|
|
Companies:
|
University of Minnesota and University of Minnesota
|
|
Address:
|
Box 807 Mayo, 420 Delaware St., S.E., Minneapolis, MN, 55455,
|
|
Keywords:
|
nonresponse bias ; survey methods ; weighting ; eligibility ; stratification
|
|
Abstract:
|
Differential nonresponse can bias results of surveys to ascertain the health or risk factors in a population. Methods have been suggested for dealing with the problem of nonresponse in epidemiologic surveys. These methods implicitly assume that all nonrespondents are part of the target population. However, contact lists for many studies contain a mixture of individuals who are and are not eligible for the study; ineligibles are weeded out by examining responses to the survey. Although the lists often contain some individual-level covariates (e.g., age and sex), identifying eligible individuals among nonrespondents may not be possible. We develop estimators of the frequency of a particular characteristic in the target population that are adjusted for nonresponse in the face of unknown eligibility. We examine several approaches to this problem using modifications to Horvitz-Thompson weighting methods that incorporate internal information on stratum-specific eligibility. We identify the assumptions necessary to make the estimators unbiased or nearly unbiased. Surveys of colorectal cancer screening behavior and of occupational violence to nurses serve to illustrate the methods.
|
