|
Activity Number:
|
501
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Thursday, August 2, 2007 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Survey Research Methods
|
| Abstract - #308779 |
|
Title:
|
Bayesian Approaches to Sequential Selection of Survey Design Protocols
|
|
Author(s):
|
James Wagner*+ and Trivellore E. Raghunathan
|
|
Companies:
|
University of Michigan and University of Michigan
|
|
Address:
|
G394 Perry, Ann Arbor, MI, 48106-1248,
|
|
Keywords:
|
Adaptive Design ; Survey Design Protocols
|
|
Abstract:
|
Surveys sequentially choose design protocols for contacting and interviewing cases. Most surveys follow very general protocols that minimally differentiate cases from each other. We develop a series of models to help identify most efficient strategies for cases conditional on their fixed characteristics and history of previous attempts. These models use priors specified from other surveys and data from the sampling frame and call records to derive posterior probabilities of contact and interview. These posterior probabilities are then used to determine the best protocol to be used for maximizing the chances of contact or interview. Results from random digit dial and face-to-face surveys analyzed from this perspective are presented.
|
