| |
 |
|
|
Science Forum Index » Statistics - Math Forum » Non stationary process feature extraction...
Page 1 of 1
|
| Author |
Message |
| unixops... |
 Posted: Wed May 07, 2008 10:04 pm |
|
|
|
Guest
|
Is there a known process for normalizing the variance of a
multivariate function using a moving average? |
|
|
| Back to top |
|
| unixops... |
| Posted: Tue May 13, 2008 5:48 pm |
|
|
|
Guest
|
On May 8, 1:04 am, unixops <clearlineofsi... at (no spam) gmail.com> wrote:
Quote: Is there a known process for normalizing the variance of a
multivariate function using a moving average?
I require process that minimizes a variance(m) and maximizes the
deviation max(1/n-1,1/n-2,..1/n-(m-1), g(x)=sum(f(x/m)-x'); m<=n.
Without a brute force computation
of every deviation. (e.g. a moving least squares fit)
? |
|
|
| Back to top |
|
| unixops... |
| Posted: Tue May 13, 2008 6:44 pm |
|
|
|
Guest
|
On May 13, 8:48 pm, unixops <clearlineofsi... at (no spam) gmail.com> wrote:
Quote: On May 8, 1:04 am, unixops <clearlineofsi... at (no spam) gmail.com> wrote:
Is there a known process for normalizing the variance of a
multivariate function using a moving average?
I require process that minimizes a variance(m) and maximizes the
deviation max(1/n-1,1/n-2,..1/n-(m-1), g(x)=sum(f(x/m)-x'); m<=n.
Without a brute force computation
of every deviation. (e.g. a moving least squares fit)
?
In regards to the above, if I have a time series where the variance
reaches a maximum for a specific window and converges to the mean at a
window greater the root mean squared difference; I could loose
information when fitting the series to a larger variance, at the same
time increase computation where the deviation converges to zero.
Therefore maximize the deviation of the variance that minimizes the
error. |
|
|
| Back to top |
|
| |
|
Page 1 of 1
All times are GMT - 5 Hours
The time now is Sun Jul 06, 2008 8:07 pm
|
|
