where v is a nX1 vector and S is a nXn sample covariance matrix of some n dimensional random vectors. I want to treat S as a matrix of random variables and get the variance of v'Sv.