[Numpy-discussion] simulate AR

[Skipper Seabold]

On Fri, Oct 14, 2011 at 2:18 PM, Alan G Isaac <alan.isaac@gmail.com> wrote:
>
> On 10/14/2011 1:42 PM, josef.pktd@gmail.com wrote:
> > If I remember correctly, signal.lfilter doesn't require stationarity,
> > but handling of the starting values is a bit difficult.
>
>
> Hmm.  Yes.
> AR(1) is trivial, but how do you handle higher orders?
>

Not sure if this is what you're after, but here I go the other way
signal -> noise with known initial values of an ARMA(p,q) process.
Here I want to set it such that the first p error terms are zero, I
had to solve for the zi that make this so

https://github.com/statsmodels/statsmodels/blob/master/scikits/statsmodels/tsa/arima_model.py#L295

This is me talking to myself about this.

http://thread.gmane.org/gmane.comp.python.scientific.user/27162/focus=27162

Skipper