Negative eigenvalues

Dear all,

I try to simulate the data using mvnrdn function but I had the the error saying that "SIGMA must be a symmetric positive semi-definite matrix". I checked the eigenvalues and the problem seems to be the negative eigenvalues of covariance matrix. Does anyone know how to fix this problem?

mean=[3.6,0.48,1.8,1.8,1.12]

var=[0.5365 0.0676 0.2527 0.2527 0.1581 
0.0676 0.0092 0.0340 0.0340 0.0213 
0.2527 0.0340 0.1321 0.1272 0.0796 
0.2527 0.0340 0.1272 0.1321 0.0796 
0.1581 0.0213 0.0796 0.0796 0.0511];

Thank you for your suggestions,

Wan

[Wayne King]

Hi Winn, I would not name your matrix, var, that is the name of a MATLAB function (variance) and that is going to cause you problems.

Is the matrix you have listed above your covariance matrix, or your variable matrix?