Handling Uncertainty in Input to Expected Value Models
Abstract
Due to the large number of entities and processes that must be represented, combat models at the theater level in the Army today are expected value models. An expected value model is deterministic-it uses the expected value of random variables as inputs and generally uses some sort of expected value within the internal processes. Use of expected value models creates problems in the proper interpretation of their output and ways for representing the uncertainty associated with the model and input and processes. This paper suggests a method for handling uncertainty in the input data sets (which usually contain elements that are specific realizations of random processes) in situations where the outcomes of interest can be expressed in binary variables (e.g. 'success' or 'failure'). A theater nuclear exchange is used as an example, having many different possible outcomes determined by random processes. A method is provided for describing the space of the exchange and partitioning the space into sets of outcomes which, if used as input into a theater-level conventional simulation, are expected to lead to significantly different results. A method for sampling the most probable outcome from each set is explained. This approach permits the construction of an experimental plan that requires a small number of model runs, each run expected to provide a significantly different result. From these runs an estimate of the variability in the theater combat resulting from uncertainty in the input data can be made.
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 01, 1989
- Accession Number
- ADA218062
Entities
People
- Mark A. Youngren
Organizations
- Center for Army Analysis