<div dir="ltr">Thanks Nikos Alexandris,<div><br></div><div style>This is helpful for me. So to get a pca plot i need to use just d.correlate with the output raster from i.pca right?</div><div style><br></div><div style>Also i.pca changes spatial position and to output of i.pca will have the changed pixel position</div>
<div style><br></div><div style><br></div></div><div class="gmail_extra"><br><br><div class="gmail_quote">On Wed, Feb 6, 2013 at 2:33 PM, Nikos Alexandris <span dir="ltr"><<a href="mailto:nik@nikosalexandris.net" target="_blank">nik@nikosalexandris.net</a>></span> wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">Rashad M wrote:<br>
<br>
> Hi,<br>
<br>
Hi Rashad :-)<br>
<div class="im"><br>
> How to create a PCA plot for two channels of a landsat image.?<br>
<br>
</div>In grass use d.correlate, in R too many options!<br>
<div class="im"><br>
> i.pca outputs a eigen values, vectors and percentage importance<br>
<br>
</div>Quoting myself :-p:<br>
<br>
--%<---<br>
The eigenvalues define proportionally the length of the axes of variation<br>
and the eigen or characteristic vectors define the direction of the<br>
variation (Ahearn and Wee, 1991). Since both the eigenvectors and the PCs<br>
only define directions, they can be arbitrarily multiplied by −1<br>
(Cadima and Jolliffe, 2009).<br>
--->%--<br>
<br>
Effectively, percentages indicate the amount of variance that has been<br>
redistributed in a Principal Component -- remember, PCs are sorted from<br>
the one that holds the largest variance to the one that holds the smallest<br>
variance.<br>
<div class="im"><br>
> Could anybody explain how to plot it?<br>
<br>
</div>You mean just a bi-variate scatter-plot?<br>
<br>
PCA is a linear transformation for multivariate data sets. The new,<br>
transformed variables (or dimensions or channels or you name them) can<br>
then be plotted the same way as any other raster map. E.g., d.histogram<br>
for single stuff, d.correlate (for a scatter-plot) and probably more.<br>
<div class="im"><br>
<br>
> Does i.pca transforms/changes pixel values?<br>
<br>
</div>Yes.<br>
<br>
Normally, one would select Landsat bands of interest, i.e. bands that are<br>
profiling "wanted" landscape features. A PCA would then transform a set of<br>
bands into something new: sorted variables in which the original variance<br>
of the data is redistributed in a way that the first Principal Components<br>
contain most of it, while the higher order transformed variables contain<br>
the smallest amounts of the original variance. Note, changes tend to<br>
appear in some of the higher order PCs. Noise, is most of the time<br>
accumulated in the last PC. And, of course, there are many and diverse<br>
uses of PCs (like compression, fusion, etc.).<br>
<br>
(So,) If you have your PCs of interest, then you can scatter-plot them in<br>
grass with <d.correlate> for example. In R, however, you can load, in<br>
theory, infinite number of dimensions (in your wording == channels) and<br>
plot really nice and fancy stuff.<br>
<br>
I have tried to clearly present the PCA concept in my work. Will send you<br>
a link and stuff of mine -- they might be useful for you to make them even<br>
better (!).<br>
<br>
Additionally, recently I have seen some very nice tri-variate PC plots in<br>
some presentation... (dunno remember now, it was certainly someone inside<br>
the GRASS GIS community!).<br>
<br>
Ah, don't forget to have a look in GRASS-Wiki (and maybe help iron the<br>
page!): <<a href="http://grasswiki.osgeo.org/wiki/Principal_Components_Analysis" target="_blank">http://grasswiki.osgeo.org/wiki/Principal_Components_Analysis</a>>.<br>
<br>
Best, N<br>
<br>
</blockquote></div><br><br clear="all"><div><br></div>-- <br><div><font face="arial, helvetica, sans-serif">Regards,<br> Rashad</font></div>
</div>