ChangeLog

Version 1.0, July 2004.
-----------------------------------------------------------------
This version handles basic univariate ARIMA modeling (actually
it just does ARMA modeling, but it allows you to difference the
data before fitting the model, which is functionally equivalent
to fitting an ARIMA model.

It also handles GARCH modeling.

Both families of models can be fit using either maximum-likelihood
estimation or Markov chain Monte Carlo methods.  The MCMC approach
is not sophisticated at this point.  10000 iterations of a chain are
run, taking posteriors to be uniform over the space of allowable
parameter values (theta, phi), and uninformative (very high
variance) over the variance of the process.

Parameters can be "held" at specific values when parameter
estimation (either MLE or MCMC) is carried out.  This is handled
within the model specification dialog boxes.

Missing values are allowed.  In data files, these are
encoded as NA.