|
Activity Number:
|
196
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Tuesday, August 10, 2004 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Survey Research Methods
|
| Abstract - #301250 |
|
Title:
|
Using the Fractional Imputation Methodology to Evaluate Variance Due to Hot-deck Imputation in Survey Data
|
|
Author(s):
|
Adriana Perez*+ and John M. Finamore and Jeffrey H. Stratton
|
|
Companies:
|
University of Texas Health Science Center, Houston and U.S. Census Bureau and U.S. Census Bureau
|
|
Address:
|
80 Fort Brown SPH RAHC Rm N.200, Brownsville, TX, 78520,
|
|
Keywords:
|
ignorability ; missing at random ; item nonresponse ; serpentine sorting ; nearest neighbor ; successive difference replication
|
|
Abstract:
|
Limited variance estimation methods exist in the statistical literature for large complex sample surveys after the imputation of missing data. Since variance estimates depend on both the sample design and the imputation method, methods under design-based or model-assisted perspectives exists. This paper examines empirically the effect on the variance estimate due to the use of hot-deck imputation with a nearest neighbor donor in the 1999 Survey of Doctorate Recipients. This examination involves a comparison to results from the pairwise fractional hot-deck imputation methodology. For this evaluation, two distinct donors were selected for each missing value and imputation fractions for the donors were computed. One hundred sixty replicate weights were created to determine variance estimates. Each replicate included an adjustment to account for the imputation fractions. The successive difference replication variance estimation method was used. In addition, three stages of weighting adjustments were evaluated to determine the most appropriate stage to incorporate the fractional imputation weight adjustment.
|
