Accident Metric (AM)¶

Description¶

AM evaluates whether an accident happened in a scenario:

\[\begin{split}\mathit{AM}(\mathit{Sc}) = \begin{cases} 0, \quad \text{no accident happened during } \mathit{Sc}, \\ 1, \quad \text{otherwise.} \end{cases}\end{split}\]

This simplistic metric is implicitly used in accident databases, such as GIDAS. It fails to identify critical non-accident scenarios.

Properties¶

Run-time capability¶

Target values¶

Not necessary

Subject type¶

Any

Scenario type¶

Any

Inputs¶

Any parameter combination that allows determination of whether an accident happened or not

Output scale¶

\(\{0,1\}\), nominal scale

Reliability¶

Low, as a change in accident outcome can be related to only minor changes in criticality (i.e. near misses)

Validity¶

Medium, due to maximal specificity but low sensitivity

Sensitivity¶

Low, as all critical non-accidents are missed

Specificity¶

Maximal as all accidents are critical

Prediction model¶

None, since a-posteriori