A Method for Simulating Correlated Random Variables From Partially Specified Distributions.
Abstract
Historical data from which to develop reliable estimates of the statistical relationships among cost elements are often scarce. Typically, information comes from many different subject-matter experts, who provide at most the moments (mean or median or mode, and possibly variance), distributional form (e.g., beta or triangular), bounds (lowest and highest possible cost), if applicable, and correlation matrix. Eliciting a fully specified multivariate distribution for the entire set of cost elements comprising a system is usually impossible. Therefore, an algorithm is needed for simulating correlated random variables from partially specified, non-normal distributions. This paper presents such an algorithm and explains how it was developed.
Document Details
- Document Type
- Technical Report
- Publication Date
- Oct 01, 1994
- Accession Number
- ADA288820
Entities
People
- Matthew S Goldberg
- Philip M. Lurie
Organizations
- Institute for Defense Analyses