[EMAIL PROTECTED] (psu) wrote in message news:<[EMAIL PROTECTED]>...
> Can it not be argued that if the noise structure is 'well known', for
> example, we know for a fact that the noise is an uncorrelated white
> noise process, then the 'optimal' filter for this particular signal
> embedded in such a noise is simply the expectation of the process at any
> time t?
> 
> i.e.,
> 
> If we assume that Observations = Signal + Noise
> 
> we can take expectations to obtain:
> 
> E(obs) = E(sig) + E(noise)
> 
> since E(noise) = 0
> 
> E(obs) = E(sig)
> 
> So, taking expectations gives us the exact structure of the signal in
> this case...
> 
> so, is it reasonable then to say that more advanced filtering methods
> are required only when the noise cannot be said to be a simple additive
> white noise process, such as produced from measurement uncertainty?
> 
> .
>Yes ...

See http://www.autobox.com/outlier.html and in particular
http://www.autobox.com/t1b3.html

For a ton of "stuff" on time series/signal/outliers see
http://www.autobox.com/teach.html ..

For free software http://www.autobox.com/freef.exe

Dave Reilly
Automatic Forecasting Systems
hrrp://www.autobox.com
.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
.                  http://jse.stat.ncsu.edu/                    .
=================================================================

Reply via email to