|
Activity Number:
|
52
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Sunday, August 8, 2004 : 4:00 PM to 5:50 PM
|
|
Sponsor:
|
Section on Survey Research Methods
|
| Abstract - #300447 |
|
Title:
|
Calibration-weighting and Its Effect on Measures of Variation
|
|
Author(s):
|
Richard A. Moore, Jr.*+
|
|
Companies:
|
U.S. Census Bureau
|
|
Address:
|
, Washington, DC, 20233,
|
|
Keywords:
|
calibration-weighting ; random group replication ; general variance function
|
|
Abstract:
|
We are taking a random sample of n units from a universe with N units. Suppose a frame variable of interest (e.g., receipts) has a highly skewed distribution. Each selected unit is assigned a weight of N/n. Since the distribution is highly skewed, it is very likely that the estimate for the aggregate value of the weighted sum of this variable over the selected cases differs significantly from the actual aggregate sum over all units in the universe. Suppose we adjust the weights of the selected units, so that (1) the adjusted weighted sum of the frame variable over the selected units equals the sum of the variable for all cases in the universe, and (2) the sum of the adjusted weights equals the number of cases in the universe. This will then require some modifications to the method used for calculating the variance. Using the employer universe of the 2002 Survey of Business Owners, this paper compares several alternative variance estimation methods and focuses on the accuracy of the variances under a calibration-weighting scheme.
|
- 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 |
