Use of uncertainty analysis in evaluation and validation

The USEPA s Committee on Regulatory Environmental Models defines model (or data) evaluation as 

Validation is the process by which the reliability and relevance of a particular approach, method or assessment are established for a defined purpose (IPCS, 2004). Uncertainty analysis can contribute to this, by indicating the reliability of the exposure estimate (how different the true exposure might be) and hence its usefulness for decision-making. 

An especially useful form of validation is where the results of an assessment can be compared with independent data or information (e.g. comparing predicted exposure with biomarker measurements or epidemiological studies). When making such comparisons, it is important to remember that both sides of the comparison are subject to uncertainty. The methods described in this document should be applied to the independent data, as well as to the exposure assessment, to provide a fair comparison between the two. 

Adequate documentation of all aspects of the uncertainty analysis for the exposure assessment is also critical to ensure transparency to reviewers and stakeholders. This includes sufficient documentation to enable independent replication of results and necessitates detailed description of qualitative and quantitative aspects pertaining to data, scenarios, methods, inputs, models, outputs, sensitivity analysis and interpretation of results. 

Transparency of documentation is also critical to adequate review of the uncertainty analysis by peers, an integral component of a defensible process for exposure and risk assessment. Selected peer reviewers must have representative expertise, which may include, but may not necessarily be limited to, analysts, exposure assessors, statisticians, etc. 

---