DIST-MC:  Do Markov chain sampling for the specified distribution.

The dist-mc program samples from the distribution specified with
dist-spec (see dist-spec.doc) using a specified series of Markov chain
operations (see mc-spec.doc).  

All the standard Markov chain operations are supported.  The default
stepsizes are all one.  These stepsizes may be multiplied by a
stepsize adjustment specified by the user, in the usual way.

For distributions specified by a single energy function, the inverse
temperature used in tempering methods may be interpreted in two ways,
as appropriate for traditional tempering method (-zero-temper) or for
annealed importance sampling.  See dist-spec.doc for details.  For
Bayesian posterior distributions, the inverse temperature is used to
multiply the log likelihood terms (see dist-spec.doc).  An inverse
temperature of zero therefore gives the prior distribution.

Dist-mc knows how to sample from this prior distribution if it has a
simple form (see dist-spec.doc).  Otherwise, if annealed importance
sampling is to be used, the -read-prior option must have been given to
dist-spec, which tells dist-mc to obtain points drawn from the prior
from an external source.  It does this by reading them from standard
input, as a free-form list of numbers in standard text form, with the
the state variables being read in the order: u0, ..., u9, u, v0, ...,
v9, v, w0, ..., w9, w, x0, ..., x9, x, y0, ..., y9, y, z0, ..., z9, z.

There are no "application specific" operations implemented by dist-mc.

For details regarding program arguments, see the generic documentation
in xxx-mc.doc.

            Copyright (c) 1995-2004 by Radford M. Neal