|
Activity Number:
|
533
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Thursday, August 10, 2006 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Survey Research Methods
|
| Abstract - #307036 |
|
Title:
|
A Comparison of Big Match and the Current NDI Record Selection System Using Artificial Data
|
|
Author(s):
|
Bryan Sayer*+
|
|
Companies:
|
Social & Scientific Systems, Inc.
|
|
Address:
|
921 Chatham Lane, Columbus, OH, 43221-2418,
|
|
Keywords:
|
National Death Index ; record linkage ; error rate
|
|
Abstract:
|
Using a frequency of sex-specific first names and middle initials and nonsex-specific last names from 22 years of the NDI (about 48 million people), we generate an artificial database that replicates the NDI structure and an artificial matching file that contains both decedents and nondecedents. The matching file also contains "altered" submission records that represent problems found in survey records matched to the NDI. We compare the results using a variety of relaxed blocking assignments in Big Match to estimated results under the current system for those records selected. We show which correct records are selected by one, the other, or both of the systems and which system works the best with which types of altered records.
|