|
Activity Number:
|
390
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Wednesday, August 9, 2006 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Survey Research Methods
|
| Abstract - #306060 |
|
Title:
|
Using Census Data to Define Estimation Areas for the American Community Survey: a Case Study
|
|
Author(s):
|
Joseph Powers*+ and Alfredo Navarro
|
|
Companies:
|
U.S. Census Bureau and U.S. Census Bureau
|
|
Address:
|
P.O. BOX 1607, COLLEGE PARK, MD, 20740,
|
|
Keywords:
|
reliability ; naive clusters ; statistical clustering ; compactness ; weighting ; estimation areas
|
|
Abstract:
|
In January 2005, the American Community Survey (ACS) expanded to sample all 3,219 counties in the U.S. and Puerto Rico. The ACS weighting and estimation methodology requires estimation areas to meet a minimum population size so that the observed sample size is big enough to produce estimates with adequate reliability. Counties below the threshold size must be grouped or clustered prior to estimation. A simple method groups the counties based on adjacency and then assess all the clusters using a predefined criterion. A better, automated algorithm was also developed. The algorithm is an iterative method that uses a set of Census long form characteristics to define a similarity index based on the Euclidean distance metric. This paper describes the naïve method, the algorithm, and a statistical assessment. The results of the two schemes are compared for Puerto Rico and Texas.
|
