On 29 Jan 2003 16:28:57 +0100, Peter Dalgaard BSA
<[EMAIL PROTECTED]> wrote in message
<[EMAIL PROTECTED]>:


>MCMC has big difficulty in generating *independent* samples. 

That's true.  For an IID sample, a rejection sampler (if you can do
the calculus) is usually more efficient.  To get a a truly IID sample
from MCMC you need to start chains at independent starting values and
take one observation from each; that's probably too slow. 

Duncan Murdoch

______________________________________________
[EMAIL PROTECTED] mailing list
http://www.stat.math.ethz.ch/mailman/listinfo/r-help

Reply via email to