|
Activity Number:
|
443
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Thursday, August 7, 2003 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Survey Research Methods
|
| Abstract - #302296 |
|
Title:
|
Alternative Weighting Adjustments for the Knowledge Networks Panel Based on Nonresponse Analyses
|
|
Author(s):
|
Rick J. Li*+ and Vicki J. Huggins and Mick P. Couper and J. Michael Dennis
|
|
Companies:
|
Knowledge Networks, Inc. and Knowledge Networks, Inc. and University of Michigan and Knowledge Networks, Inc.
|
|
Address:
|
1360 Willow Rd., Menlo Park, CA, 94025-1516,
|
|
Keywords:
|
RDD ; nonresponse ; internet panel ; panel attrition
|
|
Abstract:
|
Bias due to nonresponse in survey estimates is a function of the level of nonresponse and the magnitude of differences between nonrespondents and respondents to key questions of interest. There are several stages in the life cycle of building and maintaining a research panel where nonresponse or attrition enter and potentially introduce sample bias. At each stage of panel loss, there is some level of information on the members that were actually recruited or even attempted. Less information is available during RDD recruitment than at later stages. This paper will present alternative weighting schemes to reduce bias due to sample loss at the RDD stage and panel attrition by incorporating two new sets of research findings on each of these sources of nonresponse bias. The first is an analysis of nonresponse detectable by differences at the Census tract level between panel responders and nonresponders using the RDD sample fielded for panel recruitment. The second is an analysis of nonresponse resulting from attrition the panel. The data source is comprised of 397,000 interviews completed by former and current Knowledge Networks panel members.
|
