[SciPy-User] Trend Detection

[josef.pktd at gmail.com]

On Fri, Oct 2, 2009 at 10:40 AM, Gökhan Sever <gokhansever at gmail.com> wrote:
> Hello,
>
> Recently, I have come across a new paper published with the title:
> Deterministic versus stochastic trends: Detection and challenges [1]. I am
> planning to experimentally apply (very preferably, without re-inventing the
> wheel :) some of the techniques that they adapted for trend detection
> (parametric and non-parametric ones) on my datasets. (Ahh, I don't know how
> will it would take for me to fully grasp what this is really means:
> "including wavelet analysis, heuristic methods and by fitting fractionally
> integrated autoregressive moving average models.")
>
> Are any of the mentioned approaches available in SciPy habitat?
>
> Any comments or suggestions are appreciated.
>
> ---
> [1] : Fatichi, S., S. M. Barbosa, E. Caporali, and M. E. Silva (2009),
> Deterministic versus stochastic trends: Detection and challenges, J.
> Geophys. Res., 114, D18121, doi:10.1029/2009JD011960.
>
> --
> Gökhan

I haven't seen anything directly for this, there are bits and pieces
around that might make some of the functions pretty easy to write.

pytrix has some unit root tests (adf, not sure about Phillips Perron)
http://code.google.com/p/econpy/source/browse/#svn/trunk/pytrix

Also some of the tests should be easy to write with statsmodels, since
the regression and many statistics are available.

In econometrics, there is a large literature (mostly Perron) about
stochastic trend versus deterministic trends with structural breaks,
but I never read the details.

I don't think I have seen any of the non-parametric (incl. wavelets)
in this context, but I would worry a lot about the power of the tests
in this case.

some arma timeseries functions are also available in statsmodels,
including estimation and some work with impuls_response_functions. I
did it so far only for stationary processes, differencing or
deterministic detrending has to be done outside. (I never worked my
way through fractionally integrated processes).

Whatever you come up with, this would also be very interesting for
econometrics and scikits.statsmodels.

I looked at trend versus difference stationarity a long time ago in
school, and one recommendation also was to work in the frequency
domain, but I don't remember the details.
I don't know if any of the frequency domain functions in nitime (of
nipy origin) are useful for this.

Josef


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