How can I use mvnrnd function in simulation?

when using mvnrnd function we use mean and covariance matrix which is positive semi definite my question is:

When us开发者_StackOverflow社区ing this function in simulation the covariance matrix sometimes is not positive semi definite how can I deal with it? Using chol function? If so how?

---

Very similar discussion here

The short answer is that your input (which is output from your simulation) is not valid. This could be for a few reasons.

Without knowing your situation, I would point you to this resource to figure out why you are getting covariances that are not positive semi-definite, and the last section of that site gives a method of "side-stepping" the problem.