Generate correlated random variables in a wider domain

1 view (last 30 days)
Hi all,
I have a set of correlated data with rho = 0.0206. Each of two variable has a statistical distribution as follows (par = parameter):
WL = ('generalized extreme value',par1 = -0.4241, par2 = 0.6731, par3 = 1.4881)
Hs = ('Generalized Pareto', par1 = -0.0860, par2 = 0.2149, par3 = 1.7000)
when I ran my code to generate 100000 correlated samples, I see that data in X-axis has maximum limit of 3.15 and data in Y-axis has maximum limit of 3.4. I want to increase these limits to 6 but don't how!
Here is my code:
clear all
clc
% Reading raw data
RawData = xlsread('TrHs=1.7m.xlsx');
Hs = RawData(:,2);
WL = RawData(:,1);
n = 100000; % Read number of simulations
rho=0.0206; % Calculate correlation parameter
ro=[1 rho;rho 1];
Z = mvnrnd([0 0], ro, n);
U = normcdf(Z);
WLinv=icdf('generalized extreme value',U(:,1),-0.4241, 0.6731, 1.4881);
Hsinv=icdf('Generalized Pareto',U(:,2),-0.0860, 0.2149, 1.7000);
X = [WLinv Hsinv];
[n1,ctr1] = hist(X(:,1),50);
[n2,ctr2] = hist(X(:,2),50);
plot(X(:,1),X(:,2),'.r'); hold on; plot(WL,Hs,'.b','markersize',12);
legend({'Simulated Data','Raw Data'},'Location','northwest','fontname','times','fontweight','bold','fontsize',10)
title('Monte-Carlo Data Simulation','fontname','times','fontweight','bold','fontsize',12)
ylabel('Significant Wave Height (Hs)','fontname','times','fontweight','bold','fontsize',12)
xlabel('Water Level (WL)','fontname','times','fontweight','bold','fontsize',12)
csvwrite('DataGen1.csv',X) % Write simulated data into an Excel file

Answers (1)

Jeff Miller
Jeff Miller on 3 Jan 2020
It looks like 3.15 and 3.4 are the maximum values in the generalized extreme value and pareto distributions with the parameters that you have given. If you want to get values up to 6, then you will have to use different parameter values for these distributions. For example, you could get values up to 6 by increasing either of the first two parameters of the generalized extreme value distribution (location, scale)--or both. It is just a question of exactly what distribution parameters you want to use in your simulation.

Categories

Find more on MATLAB Report Generator in Help Center and File Exchange

Community Treasure Hunt

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

Start Hunting!