Monte Carlo Analysis

Th’ Gaussling is surfing the net looking at freeware. The possibilities are amazing. It’s like I’ve crawled up from underground onto a sunny Brazilian beach full of nubile sunbathers frolicking in the waves. It’s all just too wonderful to be true …

[sound of needle scratching across record]

So I’m new to Monte Carlo simulations. I have some economic modeling to do and it has become apparent to me that mere spreadsheet grinding isn’t enough.  To make a more convincing case I need to introduce plausible cost variances somehow by mixing and blending various kinds of guesswork distribution curves from the elements of the project. There is a particular Project Risk and Contingency Analysis software that seems reasonable to start with. It would be interesting to hear of experiences others may have had with Monte Carlo project modeling.

3 thoughts on “Monte Carlo Analysis

  1. John Spevacek

    I’ve run across “Crystal Ball” at a couple of different employers. I don’t know whether or not it has a ton of bells and whitles (let alone the cost) but the accountants at a Fortune 50 company used it when evaluting NPV of capital investments. At both companies, what has impressed me so much is not the answers that you get, but that the users (not just engineers, but accountants and mangers) really seem to understand what the program simulates: they explain it to me as if it is some new statistical development (not knowing that it dates back to the ’30s) – an unintended but positive feature of any program. You never get explanations like that about other new programs, such as Six Sigma.

    Reply
  2. gaussling Post author

    Ulam and von Neumann, I think. I vastly prefer smaller companies where folks are happy with two or three sigma. I’m looking for something cheap. Thanks for the heads up.

    Reply
  3. John Spevacek

    Oops. I did some digging and the starter version is $999. That is pretty steep.

    You could roll your own in Excel if you are working with some of the more standard distributions. Feed “RAND()” into NORMSINV or CHIINV or whatever. I’ve done it a few times on my own, but if you have more than a few distributions, it can get long. The advantage is that since you devised it, you know exactly what it’s doing.

    Reply

Leave a comment