|
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 - #307645 |
|
Title:
|
Using Test Databases To Evaluate Record Linkage Models and Train Linkage Practitioners
|
|
Author(s):
|
Michael H. McGlincy*+
|
|
Companies:
|
Strategic Matching, Inc.
|
|
Address:
|
, Morrisonville, NY, 12962,
|
|
Keywords:
|
codes ; crash ; outcomes ; probabilistic ; record ; linkage
|
|
Abstract:
|
Traffic safety research grantees develop Crash Outcome Data Evaluation Systems (CODES) by linking police and medical records. The process is complex because most records lack unique identifiers and exhibit high levels of misclassification and nonresponse. Grantees learn to develop Bayesian models that compare quasi-identifiers in order to estimate the probability that a record pair is a true link. To help teach effective modeling techniques, the CODES program uses test databases for which true link status is known. A data generator creates test databases for each grantee by simulating crash and medical events. Quasi-identifiers specific to each grantee are drawn from multinomial distributions. Case duplication, item misclassification, and item nonresponse are simulated as Bernoulli events. The generator captures clustering of data for vehicle occupants that occurs in real life.
|
- 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 2006 program |
