NCJ Number
242130
Date Published
September 2012
Length
51 pages
Annotation
This research developed simple, practical tools and guidelines for analysts, researchers, and West Virginia State repository staff in examining crime-data quality and improving the capacity of West Virginia and other States in producing more stable estimates of crime trends using State incident-based data that mirror reporting for the FBI's National Incident-Based Reporting System (NIBRS).
Abstract
The imputation methods developed reasonably estimated missing data that would otherwise go undetected and uncounted, with the potential for extending methods to smaller units of analysis and specific crime categories for large cities or regional areas. The West Virginia incident-based reporting (IBR) agency data were used to develop techniques for identifying issues of data quality and investigating imputation methods to estimate missing data based on 1 year of aggregated property and violent crime-count data. Common problems were found in the IBR data and patterned to provide guidance for classifying crime counts of zero as true zeros or missing data and developing outlier detection methods in the context of NIBRS data in order to identify irregular reporting. Two simulation studies were conducted using the IBR data to investigate imputation methods for estimating partial and non-reporting agency crime count data. Data used for simulation came from agencies that reported 12 months of data with no irregular crime count. The IBR missing value pattern was used to model missing data and the impact of imputation methods on partial reporting agencies (reporting 3-11 months of data). The second simulation explored the performance of imputation methods for non-reporting agencies (reporting 0-2 months of data) with 10 - 80 percent of entire agency data deleted. The proportions of data deleted were used to assist NIBS-certified data not at 100 percent crime or population coverage. 9 figures, 16 tables, and 27 references