PDA

View Full Version : Finding the right Gamma distribution

yayo1984
08-27-2010, 12:53 AM
Hi all,

I am a Masters Student working on a simulation model. As part of model construction I have to assign probability distributions for each parameter to be analyzed in the model. In my model some parameters were given values based on normal probability distributions which was pretty straight foward(clt.jpg). Now I am having an impossible time figuring out which distributions and which associated values (mean,alpha,beta, etc...) to use. I am having these issues probably due to the fact that I have no knowledge of statistics beyond basic probability.

My problem in particular(todo.jpg):

I want the range to be 0 - 1 and values in between should be selected based on a distribution curve like the one I have shown below

From searching on the internet, i think this is a gamma distribution. I am using Excel 2007 and trying to set the 'shape-alpha', 'stretch-beta' parameters in attempts to produce almost exactly this graph. After trying several permutations in excel I have almost given up hope. I tried other distributions as well but I think gamma works best. My choices on distributions are limited because of the simulation tool I am using (Vensim) to
BETA,BINOMIAL,EXPONENTIAL,GAMMA,NEGATIVE BINOMIAL,POISSON,WEIBULL.

Thank you

BGM
08-27-2010, 04:29 AM
Two things you may need to know:

1. For gamma distribution, it has an inifinite support (0, +\infty)
but in your case seems you only want to model a finite support (0, 1)

2. The area under the curve you shown is much smaller than 1.
For any well defined probability density function, the area under the curve
should be equal to 1.

Dason
08-27-2010, 07:15 AM
Yeah, out of your choices a beta distribution is the only one with a support of [0,1]

yayo1984
08-27-2010, 11:54 PM
I have used the hints and successfully produced the graph using a beta distribution. I guess it was my missing knowledge of stats that needed to be patched. Thanks for the help guys. :-)