On Fri, Oct 14, 2011 at 1:26 PM, Alan G Isaac <alan.is...@gmail.com> wrote: >>> Assuming stationarity ... > > On 10/14/2011 1:22 PM, josef.p...@gmail.com wrote: >> maybe ? > > I just meant that the MA approximation is > not reliable for a non-stationary AR. > E.g., http://www.jstor.org/stable/2348631
section 5: simulating an ARIMA: simulate stationary ARMA, then cumsum it. I guess, this only applies to simple integrated processes, where we can split it up into ar(L)(1-L) y_t with ar(L) a stationary polynomials. (besides seasonal integration, I haven't seen or used any other non-stationary AR processes.) If I remember correctly, signal.lfilter doesn't require stationarity, but handling of the starting values is a bit difficult. Josef > > Cheers, > Alan > > _______________________________________________ > NumPy-Discussion mailing list > NumPy-Discussion@scipy.org > http://mail.scipy.org/mailman/listinfo/numpy-discussion > _______________________________________________ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
