DIST-QUANTITIES:  Quantities defined for a specified distribution.

The plot and histogram programs used with the 'dist' programs,
dist-plt and dist-hist, have the generic facilities documented in
xxx-plt.doc and xxx-hist.doc.  In addition to the generic quantities
(see quantities.doc) and the Markov chain quantities (see
mc-quantities.doc), these programs can also reference the state
variables for the specified distribution, under the names used in
dist-spec.  Furthermore, if the distribution is a Bayesian posterior,
the following quantities are defined:

    P    Log of the prior probability of the parameters.

    l    Array of minus log probabilities for training cases, 
         obtained from the likelihood formula.  If not specified 
         to be an array, the average minus log probability over 
         training cases.

    in   Array of containing the n'th input for the training cases.

    tn   Array of containing the n'th target for the training cases.

Note that n is required for in and tn above - without it, they mean
entirely different things (see quantities.doc and mc-quantities.doc).
Note that, for historical reasons, 'P' is the log probability, but 
'l' is minus the log probability.

For example, if the distribution was specified by

    > dist-spec log "x1^2 + x2^2 + w^2"

then after dist-mc was run, the quantities x1, x2 and w could be
plotted using a command such as

    > dist-plt t x1x2w log | plot

If the distribution is a Bayesian posterior, the command

    > dist-plt t t2@0:9 log | plot

will plot target number 2 (numbered starting at 0) for the first ten
observations. 

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