1. Re: Generating Closed Skew Normal Distribution

Originally Posted by Shurid
Hi Everyone,

I have a similar problem. I want to generate multivariate Closed Skew Normal distribution. Is there any R package that does this job? Or is there any program that generates samples from Closed Skew Normal Distribution.

Any help would be appreciated.

http://www.sciencedirect.com/science...c&searchtype=a

2. Re: Generating Closed Skew Normal Distribution

I'm looking for a simple way to generate points on a "skewed normal" distribution in Excel, as defined only by three points on the curve: the 10%, 50% and 90% probability data points.

(Context: An expert projects there is a 10% likelihood hat the future value of something is, say, less than \$10. The 50% likelihood is \$20. And the 90% likelihood is the value is less than \$60. We want to now generate thousands of points along the distribution curve defined by those points, to do a simulation of the possible values as inputs to a larger financial model.)

If my 10-90 points were symmetric (e.g., \$10, \$20, \$30): I just do NORM.INV(rand(),mean,sd), where the mean is 20, and standard deviation is calculated using the z-value for a 10% probability: (20-10)/1.282=8
That will give me random points along a normal curve defined by those 3 known points.

But, usually in reality the 10-90 points are usually projected to be non-symmetric. (e.g., \$10, \$20, \$60)

There is simulation software (e.g., @RISK or Crystal Ball), that will automatically fit an appropriate distribution around my 3 input points. But I need to find a simple way to do this myself in Excel.

Therefore, I am thinking a "skewed normal" curve, where I apply some skewness coefficient (calculated from my 3 input points) to the NORM.INV formula... I haven't found any solutions in extensive web-searching (besides purchasing curve-fitting simulation software), so I wonder if I'm on the right track. Ideas?

3. Re: Generating Skewed Normal Distributions

Are you particularly attached to "skewed normal" as opposed to something like the gamma distribution or log normal?