Probability Distribution Resources

[Holy Bull]

I'm trying to build a simulator and am trying to find info on various irregular probability distributions to run some monte carlo tests. I've found pretty decent info on different types of distributions but I am trying to figure out how to formulas to program it up to get something spitting out numbers based on the distributions. Anyone know of any resources that might have stuff like this on it? Might be a reach asking here but i know some of you have created simulators [img]i/expressions/face-icon-small-smile.gif[/img]

[deepsouth]

I'm not exactly sure what you are asking, if you could give an example, I could probably be more helpful.

There is a freeware package called R that will spit out random numbers according to pretty much any distribution you want (binomial, normal, uniform, gamma, whatever else you can think of) ... and if you know what you're doing, you can get it to generate any other distribution you might want. The problem is that it takes a while to become comfortable using it, so this may not be worth the time investment.

http://lib.stat.cmu.edu/R/CRAN/

[Machiavelli]

I do my irregular distribution work quite informally, so I am afraid that I am nothing but a interested listener in this thread.

[Holy Bull]

Sorry, I probably didn't describe it well. Its been a long time since I've looked at this stuff.

Basically, I am trying to code a function, f(random number x), that will spit out results that adhere to a certain type of distribution. I've looked through some descriptions of distributions and found a few that I think might be able to achieve the shape I want (most continuous, skewed dists seem to be able to fit what I am looking for with the right #'s...something that tails off on 1 side and comes to a more abrupt halt on the other). However I don't know how to write the function without having a statistical package do it for me.

[AussieVamp2]

Bull, buy a textbook on Montecarlo. That will have some algorithms in there.

[KingOfTheSquares]

Bull, get out your copy of Excel and start playing with the stat functions...you can learn a lot from them. Go find a book in the library that explains them.

Then if you're going to get serious in the monte carlo area you ought to get Simulation Modeling and Analysis by Law and Kelton and plan to spend a number of hours getting boned up. L&K can also give you a decent feel on distributions. Pay close atttention to convergence criteria.

Get used to the fact that some phenomena have no explicit distributions but require another way of working the problem or require sampled data from the real world.

Then you need to spend some serious quality time on random (or lack thereof) number generators.

[Holy Bull]

Thanks for the help guys. I'm probably going to at first do something similar to Mach and try and come up with an informal function and just play around and see if I can backfit the data I'm working with.

One more dumb question, how do you get excel to take a series and graphically show its distribution? Was playing with it last night and I couldn't find it, and I couldn't find the CD with the stat software I used to use.

[Machiavelli]

HB:

Are you looking for a Histogram? Most versions of Excel have a pretty good Chart Wizard.

[Holy Bull]

Yeah,

Not pretty but this is what I am looking for.

[Machiavelli]

Arrange the data that you want to chart in adjecent columns. Select the data. Then Click "Insert .... Chart" and choose your style of chart.

[Holy Bull]

I'm not that rank a beginner [img]i/expressions/face-icon-small-smile.gif[/img]

What I was looking for was the histogram. It is in Data Analysis and not the standard Chat Wizard mechanism.

[AussieVamp2]

I'll agree with deepsouth though that R is an excellent statistical package if not completely user friendly from someone not used to that sort of thing

[Holy Bull]

Just want to say thanks for the help guys. I got my simulator going and even off this initial effort, I'm extremely pleased with the #'s it is
spitting out.