cwishart_variates

(rtools/cwishart_variates.m)

Function Synopsis

[B,d] = cwishart_variates(cm,looks,Nsamples);

Help text

function out = cwishart_variates(cm,looks,N);

 Gives a (p x p x Nsamples) matrix of complex Wishart variates
 with a covariance matrix CM and equivalent number of looks L.
 Often defined as having L - p degrees of freedom.

 CM should be positive definite hermitian (see www.mathworld.com)

 Note that the determinant (det) of a single look variate
 should equal 0, and that the det of an arbitrary look variate
 should be positive. However... for low looks the det is often
 complex or negative -  I'm not sure if this is due to floating
 point representation or a mistake in this method. Furthermore
 the output matrix is often not _exactly_ hermitian, again
 this is due to floating point I think...

 Source: Brute force evaluation.
 The method by Odell, P.L. and Feiveson, A.H. (1996)
 "A numerical procedure to generate a sample covariance matrix"
 Journal of the AMS 61, pp 199-203 may speed this up but I 
 couldn't get it to work with complex numbers....

 Assumes lower triangular part of cm is conjugate of upper

 Part of the Matlab Radar Toolbox v0.11 by Glen Davidson at www.radarworks.com
 Please acknowledge use of this toolbox in any publication or software.
 This header must be retained in any derivative work