# Double counting

### From Wiki

Double counting is a phenomenon which sometimes occurs in report presentation, when the structure of a report results in a single item qualifiying for inclusion within more than one distinct category. As a consequence, any totals shown in reports which contain double counting may appear not to correspond to the report detail.

It is the responsibility of any person using or presenting statistics to minimise the danger of misunderstandings which may arise from this technique.

For a detailed academic discussion of double counting, see [1].

## Contents |

## Double counting in MAST

Because all MAST reports are based on pre-calculated data cubes, **double counting is impossible in MAST**.

The main technique used to avoid double counting, while still exposing the data to rigorous analysis, is the use of boolean values such as the top level of the Crash Involved Vehicle type dimensions.

## Example of double counting

In road safety, a common example of double counting can occur when crashes are categorised by the types of vehicle involved in them. For instance, suppose that four crashes had occurred on a particular stretch of road, involving the following vehicles:

- a car and a motorcyle
- a single car
- a bus and a motorcycle
- a car and a pedal cycle

This information could legitimately be reported by asserting that there were:

- three crashes involving cars
- two crashes involving motorcycles
- one crash involving buses
- one crash involving pedal cycles

However if this report was carelessly presented, the unwary reader might jump to an incorrect conclusion: that there were *seven* crashes on the stretch of road, rather than only four. In addition, if the report showed the correct total of four, the same reader might notice that the individual values did not add up to the total and erroneously conclude that the information shown in the report was inaccurate or unreliable.

## Real life example of double counting

Double counting is sometimes a valid approach to reporting certain sorts of information. for example, table 23c of RRCGB 2008 would report a crash involving a car and a motorcycle in two different places: once in cell H29, and again in cell G38. As a consequence, the totals reported in row 83 *do not add up* to the sum of the preceding individual values.