|
Activity Number:
|
498
- Improving Data Quality and Estimation Methods for the Current Employment Establishment Survey
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Wednesday, August 2, 2017 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Social Statistics Section
|
|
Abstract #324402
|
View Presentation
|
|
Title:
|
Construction and Assessment of Generalized Variance Functions for an Establishment Survey
|
|
Author(s):
|
Kevin Roche* and Julie Gershunskaya
|
|
Companies:
|
U.S. Bureau of Labor Statistics and U.S. Bureau of Labor Statistics
|
|
Keywords:
|
Establishment survey ;
Balanced Repeated Replication ;
Generalized Variance Functions ;
R ;
Cluster Variables
|
|
Abstract:
|
The Bureau of Labor Statistics Current Employment Statistics (CES) survey leads to the creation of measures which are used to create Principal Federal Economic Indicators. One of those measures is the monthly change in establishment jobs, which CES estimates for detailed industries as well as geographic domains such as states and metropolitan areas. CES calculates sampling errors using Balanced Repeated Replication, but estimates of sampling variance can be volatile and may not be available as readily as desired. Generalized Variance Functions (GVF) are fit using regression models to existing direct estimates of sampling variance to improve estimates of those variances. A GVF should provide good fits for data used to construct the model, and it should provide good estimates of variances for other observations not used in the model. This paper develops a GVF model for ratio estimators and considers two main characteristics: accuracy in terms of confidence interval coverage, and stability. Metrics and tests for each characteristics are developed. We also consider and test new model types to further increase coverage and stability of the GVF variances.
|
Authors who are presenting talks have a * after their name.
