Stefan,

What is meant by 'spatially correlated random variables'? Correlated how? You 
could make a random image, and then smooth it and the pixels would have nice 
small scale correlations but that's not what you are asking I expect.

If you define a covariance matrix how you like - you can then diagonalise it 
with a coordinate transformation (matrix mul) - then generate a random sequence 
of vectors based on the sigma_i's - then transform back to the original 
coordinate system and your variables are correlated according to the specified 
covariance matrix. All these things can be done vectorially and efficiently.

Lognormal - well just make a normal distribution and then do 10**x !

hope that helps. If it doesn't try getting some inspiration at the Brewer's Art 
downtown. :-)

Karl



On 23/05/2012, at 12:18 PM, Stefan wrote:

> Dear All,
> 
> Do you know how to generate spatially correlated random variables with pdl?
> Lognormal distribution is preferred, but Gaussian would be acceptable.
> 
> What CPAN modules could be helpful?
> 
> I need to generate random material properties for data points in space (x,
> y, z). The properties don't change abruptly, but rather transition smoothly.
> In other words, the correlation has a finite radius, d, and then it
> disappears.
> 
> Your help will be appreciated.
> 
> Thank you,
> 
> Stefan
> 
> -------------------------------------
> Stefan Szyniszewski, Ph.D.,P.E.
> Post-Doctoral Researcher
> Civil Engineering Department
> The Johns Hopkins University
> 
> Latrobe Hall 109
> 3400 Charles St.
> Baltimore, MD 21218-2682
> Phone:  352-328-5434
> Fax:  410-516-7473
> Email: [email protected]
> Web: www.szyniszewski.com
> 
> 
> 
> 
> 
> 
> 
> _______________________________________________
> Perldl mailing list
> [email protected]
> http://mailman.jach.hawaii.edu/mailman/listinfo/perldl


_______________________________________________
Perldl mailing list
[email protected]
http://mailman.jach.hawaii.edu/mailman/listinfo/perldl

Reply via email to