|
Activity Number:
|
284
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Tuesday, August 4, 2009 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Survey Research Methods
|
| Abstract - #303622 |
|
Title:
|
Imputation Methods for Adaptive Matrix Sampling
|
|
Author(s):
|
Jeffrey Gonzalez*+ and John L. Eltinge
|
|
Companies:
|
Bureau of Labor Statistics and Bureau of Labor Statistics
|
|
Address:
|
Office of Survey Methods Research, Washington, DC, 20212,
|
|
Keywords:
|
Adaptive Design ; Burden Reduction ; Multiple Imputation ; Panel Survey ; Sample Survey ; Variance Estimation
|
|
Abstract:
|
Matrix sampling methods involve dividing a lengthy questionnaire into subsets of questions and administering each subset to subsamples of a full sample. In a panel survey, information about a sample unit can be learned during the first interview and this information can be used both to assign questions and to impute missing quantities at later interviews. Previous research has considered estimators based on available cases and simple adjustments to the design weights (Gonzalez and Eltinge 2008). Here we extend this research by developing an imputation procedure for recovering the data not collected from a sample unit at subsequent interviews. We use data from the Consumer Expenditure Quarterly Interview Survey to explore potential efficiency gains incurred from incorporating these imputation methods into the estimation procedures of an adaptive matrix sampling design for a panel survey.
|
