MATLAB Answers

generating correlated random variables

16 views (last 30 days)
Hello
Though the subject might seem similar to other subjects available in this forum, it is indeed different. I am doing a Monte Carlo simulation. In each iteration, i have to consider the spatial variability of a parameter(e.g. C) along a curve. in fact, in each iteration the curve is divided to some segments and the random variables (C1,C2, C3, ...,C8) are assigned to each of these segments. Each of these C1, C2, ..,C7 have a lognormal distribution with the same parameters and there is only correlation between C7 and C8 with a correlation matrix like
rho = [ 1,0,0,0,0,0,0,0;
0,1,0,0,0,0,0,0;
0,0,1,0,0,0,0,0;
0,0,0,1,0,0,0,0;
0,0,0,0,1,0,0,0;
0,0,0,0,0,1,0,0;
0,0,0,0,0,0,1,0.2;
0,0,0,0,0,0,0.2,1]
How can i model such a process?
it is worthy to point that, there may be different numbers of random variables in other iterations of Monte carlo simulation (i.e. due to a different curve length) but they all have a lognormal distribution with the same parameters.
Any help/idea/recommendation is highly appreciated.
Best Regards

Accepted Answer

Jeff Miller
Jeff Miller on 11 May 2021
Generate C1-C6 separately (independently) from whatever lognormal you want.
Then generate C7-C8 as a pair with the desired correlation. One easy way to do that is to first generate c7,c8 with mvnrnd and then form C7=exp(c7) and C8 = exp(c8). You just have to adjust the mu & sigma parameters of mvnrnd to get the desired distribution and correlation for C7 & C8,
  5 Comments
Jeff Miller
Jeff Miller on 6 Jun 2021
I am not familiar with that relationship/equation but it is fine if it gives you the correlation you want between C7 and C8.

Sign in to comment.

More Answers (0)

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!