| Re: normalisation of PCA bands [message #77966] |
Fri, 14 October 2011 04:51 |
lecacheux.alain
Messages: 325
Registered: January 2008
|
Senior Member |
|
|
On 14 oct, 09:59, eva.ivits-was...@ext.jrc.ec.europa.eu wrote:
> On Oct 13, 6:14 pm, alx <lecacheux.al...@wanadoo.fr> wrote:
>
>> On 13 oct, 09:58, eva.ivits-was...@ext.jrc.ec.europa.eu wrote:
>
>>> Good morning!
>
>>> How can I normalisePCAbands so that they have values between -1 and
>>> 1? Is it a simple min-max scaling or is there a more sophisticated,
>>> correct way to do it?
>
>>> Thanks in advance,
>>> Eva
>
>> You may use the STANDARDIZE function which normalizes the variance and
>> subtracts the average of each band.
>> alx.
>
> Hi Alex,
>
> Thanks but I'm afraid it isn't that I'm looking for.
>
> In meteorological publications I've read the following statement:
> "the spatial patterns (eigenvectors) properly normalized (divided by
> their Euclidean norm and multiplied by the square root of the
> corresponding eigenvalues) are called loadings; they represent the
> correlation between the original data (the time series) and the
> corresponding principal component time series."
>
> Doing this the PCA bands will have values between -1 and 1.
> Unfortunately I do not really understand the meaning of the above
> sentence...
>
> Any idea?
>
> Thanks,
>
> Eva
Hi Eva,
> Doing this the PCA bands will have values between -1 and 1.
Why ?
I am afraid to do not have fully understood what you (or the
meteorologists) mean by "PCA bands". In general, and regarding the
data you process by using the PCA method, normalizing standard
deviations of your observed variables is more or less equivalent to
normalize the eigenvalues of their covariance matrix and then to
"equilibrate" the relative importance of each variable.
Sorry to do not be a meteorologist...
alain.
|
|
|
|
| Re: normalisation of PCA bands [message #77968 is a reply to message #77966] |
Fri, 14 October 2011 00:59  |
eva.ivits-wasser
Messages: 8
Registered: October 2011
|
Junior Member |
|
|
On Oct 13, 6:14 pm, alx <lecacheux.al...@wanadoo.fr> wrote:
> On 13 oct, 09:58, eva.ivits-was...@ext.jrc.ec.europa.eu wrote:
>
>> Good morning!
>
>> How can I normalisePCAbands so that they have values between -1 and
>> 1? Is it a simple min-max scaling or is there a more sophisticated,
>> correct way to do it?
>
>> Thanks in advance,
>> Eva
>
> You may use the STANDARDIZE function which normalizes the variance and
> subtracts the average of each band.
> alx.
Hi Alex,
Thanks but I'm afraid it isn't that I'm looking for.
In meteorological publications I've read the following statement:
"the spatial patterns (eigenvectors) properly normalized (divided by
their Euclidean norm and multiplied by the square root of the
corresponding eigenvalues) are called loadings; they represent the
correlation between the original data (the time series) and the
corresponding principal component time series."
Doing this the PCA bands will have values between -1 and 1.
Unfortunately I do not really understand the meaning of the above
sentence...
Any idea?
Thanks,
Eva
|
|
|
|
| Re: normalisation of PCA bands [message #77977 is a reply to message #77968] |
Thu, 13 October 2011 09:14 |
lecacheux.alain
Messages: 325
Registered: January 2008
|
Senior Member |
|
|
On 13 oct, 09:58, eva.ivits-was...@ext.jrc.ec.europa.eu wrote:
> Good morning!
>
> How can I normalise PCA bands so that they have values between -1 and
> 1? Is it a simple min-max scaling or is there a more sophisticated,
> correct way to do it?
>
> Thanks in advance,
> Eva
You may use the STANDARDIZE function which normalizes the variance and
subtracts the average of each band.
alx.
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
