Contact

GlossaryPublic records, privacy, and data quality

Aggregation

Aggregation combines records into totals or other summaries. See how grouping rules, counting units, and denominators affect the result.

Aggregation combines individual records or smaller groups into a summary. A dashboard might aggregate incidents by month, neighborhood, or call type. The result depends on both the grouping rule and the calculation: counting incidents, adding durations, and averaging response times produce different kinds of summaries.

What a summary includes

A monthly total brings many records together under one period. A citywide total brings places together under one geography. These summaries make patterns easier to see, but the underlying differences become less visible. The National Forum on Education Statistics describes the same distinction between combining data and examining smaller groups; the principle also applies to public-safety charts. Forum guide presentation, pages 6 and 10.

Start by identifying the counting unit. A total of rows is only a total of incidents if each included row represents one incident. If a record set has one row per offense, an incident involving multiple offenses can contribute more than one row.

Example

Hypothetical: A department records 40 incidents in January, 35 in February, and 45 in March. If each incident belongs to exactly one month and the inclusion rules are unchanged, the quarterly aggregate is 120 incidents.

Now suppose the department reports a 10-minute average response time for a month with 10 responses and a five-minute average for a month with 90 responses. The combined average is 5.5 minutes: (10 × 10 + 5 × 90) ÷ 100. Simply averaging the two monthly averages would give 7.5 minutes and weight the small month too heavily.

Common confusion

Adding categories works only when the categories fit the intended total. If incidents can belong to several displayed groups, adding the groups may count an incident more than once. A dashboard should make overlapping categories clear.

Averages, percentages, and rates usually need their underlying counts or weights to be combined correctly. A median cannot generally be reconstructed from a list of group medians. A citywide percentage is not automatically the simple average of neighborhood percentages.

Reading an aggregate responsibly

Look for the period, geographic boundary, included categories, and treatment of missing values. A large total can be useful for describing overall workload while still hiding differences among times or places. Examine a relevant breakdown before using the total to describe every part of a community.

Where a summary excludes some records, the label should describe the included set. “Incidents with valid locations” is more precise than “all incidents” when unmapped records are absent.

Sources