mad anthony

Rants, politics, and thoughts on politics, technology, life,
and stuff from a generally politically conservative Baltimoron.

Sunday, December 05, 2004

Help, I need somebody, help...

I'm currently taking a class called "Analysis and Decision Making", which consists of doing various simulations using Excel and some plug-ins from a company called Palisades Decision Tools. So far I'm doing OK in the class, despite the fact that I'm not real confident in what I'm doing...

I'm trying to do a somewhat complicated analysis for a project I have due, well, tomorrow. It uses a plugin called @Risk. The simulation involves a company selling two items, with possibly correlated demands that are measured using a function called RISKCORRMAT. If that made no sense to you, it didn't make any to me either.

So I finally get what I think is a properly constructed model, go to run the simulation, and get the following great error message: The correlation matrix "[cables2.xls]Sheet1!k6:l7 specifies an invalid column(0).

So I go to @Risk help, and do a few searches, until I finally find the help paragraph, which explains everything in a clear, easy to understand manner. Actually, no. This is what it said:

invalid matrix specifies inconsistent simultaneous relationships between three or more inputs. It is quite easy to make a correlation matrix which is invalid. A simple example is: correlate input A and B with a coefficient of +1, B and C with a coefficient of +1, and C and A with a coefficient of -1. This example is clearly illegal, but invalid matrices are not always this obvious. In general, a matrix is valid only if it is positive semi-definite. A positive semi-definite matrix has eigenvalues which are all greater than or equal to zero, and at least one eigenvalue that is greater than zero.

It goes on from there, but I don't want to put you to sleep.

I work in tech support, and I regularly want to scream when I'm dealing with people who aren't too clear on the difference between left and right click. I know that I don't always explain things as clearly as I should, but now I don't feel so bad. Tip to the @Risk help writers - try to avoid using terms like clearly, obviously, and simple when you are talking about something that most people who don't find simple, clear, or obvious. Chances are, if they do, they don't need the freakin' help file in the first place.

Also, try not to use words like eigenvalue without defining them. Ask yourself: Would my mom know what this word means? If the answer is no, define it (you may need a different question if your mom happens to be a mathmetician).

BTW, I figured out what I did wrong - I had a 1 in one of the formulas that should have been a 2. What the formula actually does I'm not as clear on...


Post a Comment

<< Home