|
Activity Number:
|
252
- SPEED:Improving Survey Data Quality with Multiple Data Sources, Administrative Data, and Nonresponse Bias Control, Part 2
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, July 29, 2019 : 2:00 PM to 2:45 PM
|
|
Sponsor:
|
Survey Research Methods Section
|
|
Abstract #307621
|
|
|
Title:
|
Calibration Weighting for Nonreporting Agencies in FBI’s National Incident Based Reporting System
|
|
Author(s):
|
Philip Lee* and Dan Liao and Marcus Berzofsky and Alexia Cooper
|
|
Companies:
|
RTI and RTI International and RTI and Bureau of Justice Statistics
|
|
Keywords:
|
Administrative Data;
Crime Statistics;
Sample Representativeness;
Auxiliary Data;
Generalized Exponential Model
|
|
Abstract:
|
One of the barriers to using administrative data for official statistics is coverage error due to differences between the target population and the set of units who contribute to the administrative data. This is the case with the FBI’s National Incident Based Reporting System (NIBRS). NIBRS is used to produce estimates of crime and arrest in the U.S., but only has participation from 40 percent of law enforcement agencies (LEAs). Therefore, proper statistical adjustment is needed to align the distributions of the NIBRS reporting LEAs with the distribution of the population LEAs when generating national estimates with the NIBRS data. In this presentation, we demonstrate a calibration weighting approach that utilizes several auxiliary data sources and employs functions in the R “sampling” package to adjust for nonreporting LEAs in NIBRS. Regression tree methodology is used to select predictors in the calibration model among candidate variables that might be associated with reporting status or with key characteristics of crime. The proposed methodology can be tailored for other administrative data.
|
Authors who are presenting talks have a * after their name.
