Data Quality Examples

Some basic quality control routines are demonstrated below. Each example discusses and provides the SPSS code determining which of the records in the data set do not meet the criteria. These sample routines can be adapted easily for use at the state or local level to evaluate data quality (before submission to the FBI) or employed by analysts who want to assess the quality of any NIBRS or IBR dataset. Some of these routines can be expanded to include checking for specific data values in addition to checking for the presence or absence of data or data segments. Expanding these examples in this manner will circumvent the need for comparing output lists to FBI guidelines and/or internal coding rules. This is very helpful for systems that do not include look-up tables or built-in validation tables. Also, the SPSS code used in these examples can be modified for similar quality control needs involving different data segments and/or different variables. If you need assistance with any of the examples presented below, please contact JRSA.

Additional data quality checks and examples can be found in An Examination of NIBRS Elements.

  1. Ensure that required fields contain data. According to the FBI Data Collection Guidelines, each incident record must have at least one offense segment, one victim segment, and one offender segment This example shows how to determine which of the incident records comprising a single agency's data set fail to meet these criteria.

  2. Ensure that the required homicide (09A, 09B, 09C) offense segment data fields are complete. For incidents that involve the homicide offenses of murder and nonnegligent manslaughter, negligent homicide, or justifiable homicide, are the required offense segment data fields completed and are the values valid? For homicide offenses, the FBI requires additional offense variables of UCR offense code, offense attempted or completed, and type of weapon/force involved. This example shows how to determine which of the incident records comprising a single agency's data set fail to meet these criteria for the offense segment records.

  3. Ensure that the required homicide (09A, 09B, 09C) victim segment data fields are complete. For incidents that involve the homicide offenses of murder and nonnegligent manslaughter, negligent homicide, or justifiable homicide, are the required victim segment data fields completed and are the values valid? For homicide offenses, the FBI requires additional victim segment data fields: type of victim, related offender number, relationship of victim to offender, aggravated assault/homicide circumstances, and additional justifiable homicide circumstance (the latter for justifiable homicide offenses only). This example shows how to determine which of the incident records comprising a single agency's data set fail to meet these criteria for the victim segment records.

  4. Ensure that offenses coded as occurring at midnight are correct. Does the number of incidents supposedly occurring during the midnight hour seem reasonable given the data set under consideration? According to the FBI Data Collection Guidelines, the incident date/hour variable represents the month, day, year, and hour when the incident occurred or started, or the beginning of a time period, if appropriate. Zero, which is supposed to represent the midnight hour, might be mistakenly used to report missing or unknown data. The FBI does not screen submissions for this coding error. This example shows the proper use of the "zero." It also provides some alternatives for recoding the incident hour for use in analyses if the results of the quality control exercise suggest a large volume of incident records are affected.

  5. Ensure that victim variables are reported where required and are correct. For those offenses for which a victim injury, a related offender number, or a victim-offender relationship is required, are these variables reported on the victim segment? For nine offenses against persons (100, 11A, 11B, 11C, 11D, 120, 13A, 13B, and 210), the FBI coding rules require the reporting of the type of victim injury on the victim data segment. For all crimes against persons and robberies, the related offender number and the victim-offender relationship variables on the victim data segment are required. This example shows how to determine which of the incident records comprising a single agency's data set fail to meet these criteria for the victim segment records.