The DIST, Distribution String: A Digital Representation for Probability Distributions

Sam L.Savage
savage@stanford.edu
Consulting Professor
Management Science & Engineering
Stanford University


Wednesday, October 29, 2008
4:30 - 5:45 PM
Terman Engineering Center, Room 453


Abstract:

The DIST is a compressed string of Monte Carlo trials. It may be stored in the single cell of a spreadsheet or element of a data base, and has been used in the field at Shell Exploration & Production, and Merck & Co. An industry standard DIST 1.0 Format has just been established by a consortium including Oracle, SAS, Frontline Systems, and AnalyCorp.

If properly generated, the DISTs of two random variables, X and Y have the following properties.

1) DIST(X+Y) = DIST(X) + DIST(Y)

and

2) DIST(F(X,Y)) = F(DIST(X),DIST(Y))

These properties allow simulation results to be consolidated across an enterprise and audited.

The DIST data type is now supported by several software packages, which will be demonstrated along with applications.





Operations Research Colloquia: http://or.stanford.edu/oras_seminars.html