Multivariate lognormal distribution

[Przemyslaw Sliwa]

Hi folks,

I have a slight problem. I must simulate a multivariate lognormal
distribution with a given covariance matrix C. What should I do.
My idea was first to simulate the multivariate normal distribution, say X
and then take a new vector Y=exp(X), which has then the multivariate
lognormal distribution with the covariance matrix C.

Is my idea correct?

And does anybody know, how to estimate the covariance C_hat with a given
set of observations (lognormally distributed)?

Thanks for help,

Przem