## Friday, July 2, 2010

### A note about cost risk analysis

Dave Hulett has a whitepaper on his web site that covers an introduction to cost risk analysis.  If this is a new subject for you, this whitepaper goes through a number of quantitative considerations, including correlation risks among cost elements, that will give you a good feel for this territory.

You might also take a look at a more thorough treatment by reading through the cost estimating guide from the General Accounting Office (GAO), an arm of Congress.  One thing about this guide is that is well illustrated with a lot of diagrams that point the way.

Of course, NASA has a readable manual on parametric cost estimating.  Take a look at it if you really want to dig deeper into the subject.

Bookmark this on Delicious

1. In fact the classical central limit theorem cannot explain the asymmetrical form of the risk functions. From one side the number of units in the chain of serial temporal events is big enough to satisfy the theorem and be symmetrical, but from another side as a rule it is asymmetrical.
All abstruse methodologies and articles on this subject are unable to give an answer to this fundamental contradiction.
All of the above approaches do not solve the problem, but circumvent it.
Using the same method of Monte Carlo it is possible to get anything you want. Fortunately there are many control parameters there. So you can always make what you want to see as a result.
Pavel Barsegyan

2. We must distinguish between practical techniques to overcome the difficulties of cost risk analysis from the principal issues, connected with central limit theorem.
In fact the classical central limit theorem cannot explain the asymmetrical form of the risk functions. From one side the number of units in the chain of serial temporal events is big enough to satisfy the theorem and be symmetrical, but from another side as a rule it is asymmetrical.
All abstruse methodologies and articles on this subject are unable to give an answer to this fundamental contradiction.
All of the above approaches do not solve the problem, but circumvent it.
Using the same method of Monte Carlo it is possible to get anything you want. Fortunately there are many control parameters there. So you can always make what you want to see as a result.
Pavel Barsegyan
P.S. I am not sure I have posted my first comment correctly.