Metric selection

Central Tendency Metric Selection 364 Metrics Other than Central Tendency 368... 

OCCUPATIONAL AND RESIDENTIAL RISK ASSESSMENT 371 AOELs Versus MOEs 371 Route Considerations 372 Uncertainty and Safety Factor Selection 372 Aggregation and Cumulative Risk Assessment 372 CO-OPERATIVE REGULATORY ACTIVITIES 373 SUMMARY AND CONCLUSIONS 374 Terminology 374 Framework 374 Data Requirements 374 Methodological Guidance 375 Development and Utility of Databases 375 Modeling Initiatives 375 Data Analysis 375 Metric Selection 376 Research Needs 376 Exposure Mitigation 376 Risk Assessment 376 REFERENCES 376... 

Selection of the appropriate metric for use in the risk assessment can have a significant impact on the outcome of an assessment. Regardless of the mathematical approach used, it is important that harmonized guidance exists for metric selection. Guidance should be based on a number of considerations, including data quality, data quantity, distribution type, duration of exposure and nature of the toxicology endpoint. Based on personal experience with numerous deterministic exposure assessments, it is clear that regulatory authorities differ with respect to input and output metric selections. 

Consider the operations that will be monitored when selecting metrics. Selected metrics should reflect the most likely problem areas that may arise for the specific operating situation and the available staff to maintain or monitor metrics. Techniques are described and recommended in various publications, such as the U.K. Health and Safety Executive (HSE) publication HSG254 (see Section 4.5.3). Three hypothetical examples are described below ... 

One documented method uses process safety barriers identification for metrics selection. This concept uses a combination of lagging and leading indicators associated with process safety barriers and incident escalation controls to evaluate the process safety system performance. The basis for this method is documented in the U.K. Health and Safety Executive (HSE) publication HSG254 and illustrated by Figures 4.1-4.3. The strength of this technique arises from using the combination of indicators that provides multiple perspectives for judging the surety of a barrier or escalation control. For example, this basic concept was adopted and modified by BP to focus upon three information sources to assess key control barriers as summarized below ... 

A fundamental concept in clustering is the understanding of the importance of the metric, which defines the similarity (or distance) between samples. There are numerous metrics, and depending on the metric selected, results of analyses may vary dramatically. The choice of metric normally depends on the question that is being asked of the data. 

The metrics selected for each requirement should be considered carefrilly. They should be the common measures for simple characteristics. The difficulty of making the measurements must be considered. These specifications could easily turn into quality assurance requirements in the product. 

Metrics selected should not simply focus on a single reaction step in isolation, but also consider upstream and downstream options. 

Irrespective ot the system performance metric selected, it should be evaluated tor all system damage responses that exist tor each scenario earthquake. The final calculations tor loss estimation should include appropriate statistical methods to make use ot the foil distribution ot impacts that this generates. 

---