|
Activity Number:
|
327
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Wednesday, August 11, 2004 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Survey Research Methods
|
| Abstract - #300712 |
|
Title:
|
Individual-level Multiple Systems Estimation: Predicting Census Day Residency
|
|
Author(s):
|
Elizabeth A. Stuart*+ and Alan Zaslavsky
|
|
Companies:
|
Harvard University and Harvard Medical School
|
|
Address:
|
Dept. of Statistics, Cambridge, MA, 02143,
|
|
Keywords:
|
administrative records ; multiple systems estimation ; census enumeration ; SOUP prior ; Bayesian hierarchical model
|
|
Abstract:
|
Administrative records are a promising data source for census enumeration, but a major challenge is determining which administrative records represent Census Day residents, given the interval between the administrative record dates and Census Day. Using a Bayesian hierarchical model of migration and observation in the record systems (including the census), we predict census day residency for each individual and generate estimates of Census Day population size. We extend multiple systems estimation by modeling individual-level behavior and utilizing the full information in the record systems, including covariate and date information. Using the single observation unbiased prior (SOUP) of Meng and Zaslavsky (2002) we also predict the number of unobserved individuals. Results are presented using simulated data, with file types and coverage rates based on StARS 2000 and Census 2000 files; the file systems simulated are census records as well as the StARS 2000 record systems with reliable date information. Coverage rates of Census Day population size and file coverage rates are compared when the unobserved individuals are and are not utilized in the Census Day population estimates.
|
