Ans. Use mvnrnd() function.This function takes mean (vector of 1xk), covariance (matrix of k x k), number of points (n). The output is an nxk matrix which corresponds to the multivariate normal distribution with the specified mean, covariance.
>> n=1000; mu=[-2,2]; sigma=[1 0.5; 0.5 1]; X = mvnrnd(mu, sigma, n); >> size(X) ans = 1000 2 >> mean(X) ans = -2.0350 2.0157 >> cov(X) ans = 0.9950 0.4898 0.4898 0.9894 >> plot(X(:,1), X(:,2), '.') >> grid
The problem is underspecified if only the correlation matrix is known. In this case, set the mean to a zero vector, covariance to the given correlation matrix.
Tested in MATLAB 220.127.116.119 (R2012a), Octave 3.6.2