|
Activity Number:
|
76
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Monday, August 9, 2004 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Survey Research Methods
|
| Abstract - #302188 |
|
Title:
|
Evaluation of Variance Estimation Methods Based on a SPD Longitudinal File
|
|
Author(s):
|
Smanchai Sae-Ung*+
|
|
Companies:
|
U. S. Census Bureau
|
|
Address:
|
4700 Silver HIll Road, Room 3725-3, Mail Stop 8700, Suitland, MD, 20747,
|
|
Keywords:
|
variance estimation ; replicate weights ; balanced repeated replicate ; successive difference replicate ; residual linearization ; complex survey
|
|
Abstract:
|
The Survey of Income and Program Participation (SIPP) traditionally uses Fay's balanced repeated replication method for direct variance estimation, while the Current Population Survey (CPS) uses the successive difference replication method. The DataFerrett system (through the U.S. Census Bureau website) may use the residual technique of linearization as outlined by Deville for variance estimation in the future due to its expediency in interactive computation. We evaluate variance estimates produced by these three methods for key characteristics of the Survey of Program Dynamics (SPD) based on one of its longitudinal files. We determine the differences in the variance estimates produced by the three methods and the causes of the differences, and assess the effects of the differences on statistical inferences of the SPD's key characteristics.
|
- The address information is for the authors that have a + after their name.
- Authors who are presenting talks have a * after their name.
Back to the full JSM 2004 program |
