|
Activity Number:
|
167
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 3, 2009 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Survey Research Methods
|
| Abstract - #303826 |
|
Title:
|
Evaluation of Randomization-Based Estimation and Inference Methods
|
|
Author(s):
|
Randall K. Powers*+ and John L. Eltinge
|
|
Companies:
|
Bureau of Labor Statistics and Bureau of Labor Statistics
|
|
Address:
|
2 Massachusetts Ava NE, Washington, DC, 20212,
|
|
Keywords:
|
Design optimization ; Fixed nonresponse ; Gold standard ; Main-effects model ; Responsive design ; Subsampling
|
|
Abstract:
|
In many surveys, field procedures address nonresponse with a combination of callback efforts and changes in the mode of data collection. To analyze the resulting data, one generally needs to account for the relevant features of the underlying population, the sample design, and the nonresponse follow-up plan. This paper reviews and extends some standard randomization-based approaches to such analyses, with primary emphasis on population-level estimating equations that account for (1) group membership determined by nonresponse status; (2) random assignment of sample units to specific callback plans; and (3) parameter-identification restrictions. This approach leads to relatively simple estimators for population means and for parameters related to callback patterns and collection modes. The paper closes with a detailed simulation-based evaluation of the properties of these estimators.
|
