Model Documentation is the formal (official, approved) collection of documents and data that provides detailed explanation of the rationale, assumptions, derivations, tests and other analyses that support the use of a Risk Model for a given purpose.
Documentation bridges the gap between an actual programmed model and the original Abstract Risk Model and risk model developer's thought process and justification. Model documentation is a key element in enforcing good Model Governance and is an indispensable tool for Independent Model Validation.
Why is documentation important? Documentation is the next best alternative to continuous access to a modeler's (quantitative developer's) thought process. A deployed model consists of several distinct components
- Conceptual mathematical representations (Abstract Risk Model)
- The more precise Model Specification, the concrete realization of a model that includes all necessary information for implementation
- Programmed versions (in various digital form)
- The live Model Instance that can be deployed in operation
The modeler's perception of the model is the richest and most relevant representation of the model, as it expresses the intention. Programmed versions on the other hand, are the "legal" manifestations of a model which are put in production, support decisions etc. etc. Concrete implementations are always an approximation to the intended model.
Issues and Challenges
- No public standards for what constitutes adequate model documentation