TODO for scipy.stats.models
===========================

In converting the bspline.so from a weave build to a C extension, we
found several things that should be fixed or looked into more
thoroughly.  Hopefully we can dedicate some time to this effort at the
Scipy Conf 2008.  However, many of these items should be addressed
before stats.models goes into a release of scipy.

Items
-----

* Run pychecker on the stats.models and fix numerous errors.  There
  are import errors, undefined globals, undefined attrs,
  etc... Running the command below in stats/models produced 140+
  errors.::

    # Run pychecker on all python modules except __init__.py
    $ grind "[a-z|_][a-z]*.py" | xargs pychecker

* Address the FIXME issues in the code.

* Determine and cleanup the public API.  Functions/classes used
  internally should be private (leading underscore).  Public functions
  should be obvious and documented.  Packaging should be reviewed and
  cleaned up.

* Update documentation to scipy standards.  Especially adding example
  sections showing how to use the public functions.

* Tests!  Robust tests are needed!  Of the subset of tests we looked
  at, most only checked attribute setting, not the results of applying
  the function to data.

* Remove code duplication.  smoothers.py and bspline.py define
  SmoothingSpline class.
