I am wondering if amyone could help. I am working on data assimilation at CSU where we have a cost functional that is derived from mulitvariate Bayesian framework. The problem is that this framework assumes at the moment that the errors that we use are Gaussian. We know that some of our models have non-Gaussian errors and so we wish to redesign the DA scheme with this in mind. The problem is this, we wish to have a framework that has both Gaussian and non-Gaussian errors. We define a vector Z to be the union of two other random vectors X and Y where X~MN(0,B) and Y~MLN(0,D) where B and D are covariance matrices. My question is can we define the pdf of Z?
thanks for your time
steve fletcher
------------------------------------------------------------------------------------------------------------------------------------------
Dr Steven Fletcher
Postdoctoral Fellow
Cooperative Institute for Research in the Atmosphere
Colorado State University
1375 Campus Delivery
Fort Collins, CO 80523-1375
Ph: (001) (970) 491-8376
Fax: (001) (970) 491-8241
E-mail: [log in to unmask]
------------------------------------------------------------------------------------------------------------------------------------------
|