[SciPy-User] GARCH estimation

[josef.pktd at gmail.com]

On Mon, Apr 4, 2011 at 9:59 AM, Wes McKinney <wesmckinn at gmail.com> wrote:
> On Fri, Apr 1, 2011 at 2:48 PM, Alan G Isaac <aisaac at american.edu> wrote:
>> I see Enthought covers "estimating volatility using GARCH"
>> in their Python for Quants course:
>> http://www.enthought.com/training/python_for_financial_analysis.php
>>
>> Are the GARCH estimation routines available / part of SciPy?
>>
>> Thanks,
>> Alan Isaac
>> _______________________________________________
>> SciPy-User mailing list
>> SciPy-User at scipy.org
>> http://mail.scipy.org/mailman/listinfo/scipy-user
>>
>
> I am not aware of anything in SciPy. Josef has put in some work on it
> in statsmodels:
>
> http://bazaar.launchpad.net/~scipystats/statsmodels/devel/view/head:/scikits/statsmodels/sandbox/tsa/garch.py

This needs considerable updating. It was written as the first MLE
models for time series analysis and I was more interested in the
general framework than producing a GARCH model, and a year ago we
didn't have AR and ARMA yet. (AR and similar in that module are
obsolete by now. It also still depends on numdifftools.)

It works reasonably well in "nice" examples, but the two main things
missing are parameter restrictions and analytical derivatives.

A start for generic derivatives for the main distributions, normal and
t distribution, are in
http://bazaar.launchpad.net/~scipystats/statsmodels/devel/view/head:/scikits/statsmodels/sandbox/regression/tools.py

Josef

> _______________________________________________
> SciPy-User mailing list
> SciPy-User at scipy.org
> http://mail.scipy.org/mailman/listinfo/scipy-user
>