On Tue, 28 May 2002, Ad Feelders wrote:
> Dear list,
>
> Can anybody give me pointers for the following problem?
>
> I would like to generate random draws from a high (say 25) dimensional
> multivariate normal distribution,
> where the covariance/correlation matrix should be generated at random as
> well. Since the covariance matrix
> should be positive definite, a fully random matrix is not appropriate.
>
> Is there any possibility if generating a "random" positive definite
> covariance matrix?
Yes. The simplest is to use a Wishart distribution: see any good book on
simulation. And the simplest way to do that is as the covariance matrix
of a multivariate normal sample.
--
Brian D. Ripley, ripley@stats.ox.ac.uk
Professor of Applied Statistics, http://www.stats.ox.ac.uk/~ripley/
University of Oxford, Tel: +44 1865 272861 (self)
1 South Parks Road, +44 1865 272860 (secr)
Oxford OX1 3TG, UK Fax: +44 1865 272595
|