R help - Jan 2007 - any implementations for adaptive modeling of time series?

Hallo,

my noisy time series represent a fading signal comprising of long
enough parts with a simple trend inside of each such a part.
Transition from one part into another is always a non-smooth
and very sharp/acute. In other words I have a piecewise
polynomial noisy curve asymptotically converging to the
biased constant, points between pieces are non-differentiable.

I am looking for implementations of models adequate for such a
data. Are there any possibilities to adapt the ARIMA or
MCMC?

Many thanks in advance for any help/URLs

Hi Peter,

generally speaking, wavelets are known to be good at extracting signal from
noisy data and are
adaptive but I am not familiar with any R implementation of wavelets.
A simple way of looking at changes would be to use CUSUM (strucchange
package).
I hope this helps.
Ansel.

On 1/30/07, Peter Nimda <p.nimda@gmail.com> wrote:
>
> Hallo,
>
> my noisy time series represent a fading signal comprising of long
> enough parts with a simple trend inside of each such a part.
> Transition from one part into another is always a non-smooth
> and very sharp/acute. In other words I have a piecewise
> polynomial noisy curve asymptotically converging to the
> biased constant, points between pieces are non-differentiable.
>
> I am looking for implementations of models adequate for such a
> data. Are there any possibilities to adapt the ARIMA or
> MCMC?
>
> Many thanks in advance for any help/URLs
>
> ______________________________________________
> R-help@stat.math.ethz.ch mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide
> http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
>
	[[alternative HTML version deleted]]

Hi Ansel,

thank you for the response
> generally speaking, wavelets are known to be good at
> extracting signal from noisy data and are adaptive but
> I am not familiar with any R implementation of wavelets.
I used wavelets before.

1. wavelets is just a particular case of orthogonal function
systems. playing with different orthogonal function
systems one can find some "high effective" basis.
However this approach is hardly something what one
can adapt on fly while processing the real data.

2. wavelets are not stochastic and adapting a noise
model as well as random changes in trend is also
quite essential in my case.

> A simple way of looking at changes would be to use CUSUM (strucchange
> package).
> I hope this helps.
> Ansel.
thank you Ansel, i will look at this package right away.

kind regards
--
P.


> On 1/30/07, Peter Nimda <p.nimda at gmail.com> wrote:
> >
> > Hallo,
> >
> > my noisy time series represent a fading signal comprising of long
> > enough parts with a simple trend inside of each such a part.
> > Transition from one part into another is always a non-smooth
> > and very sharp/acute. In other words I have a piecewise
> > polynomial noisy curve asymptotically converging to the
> > biased constant, points between pieces are non-differentiable.
> >
> > I am looking for implementations of models adequate for such a
> > data. Are there any possibilities to adapt the ARIMA or
> > MCMC?
> >
> > Many thanks in advance for any help/URLs
> >
> > ______________________________________________
> > R-help at stat.math.ethz.ch mailing list
> > https://stat.ethz.ch/mailman/listinfo/r-help
> > PLEASE do read the posting guide
> http://www.R-project.org/posting-guide.html
> > and provide commented, minimal, self-contained, reproducible code.
> >
>
>