X
PrevPrev Go to previous topic
NextNext Go to next topic
Last Post 10 Oct 2016 06:23 AM by  Bisma Qazi
PCA
 2 Replies
Sort:
You are not authorized to post a reply.
Author Messages

Bisma Qazi



New Member


Posts:16
New Member


--
10 Oct 2016 06:23 AM
    Firstly,I would like to ask that why am I unable to access my previous posts. Secondly, I performed PCA on an image in ENVI and also in ERDAS. There was a huge difference between Eigen values and Eigen vector matrix computed in the two platforms. Why is it so? Also, do the rows of the eigen vector matrix represent different PCS or original bands in ENVI. Say, for example, if I got a matrix like this Eigenvector Band 1 Band 2 Band 3 Band 4 Band 1 0.587621 0.598896 0.543896 0.014209 ~ Does it correspond to PC1? Band 2 0.154339 0.068470 -0.217023 -0.961453 PC2? Band 3 0.294615 0.438520 -0.807980 0.260903 PC3? Band 4 0.737619 -0.666585 -0.065164 0.085646 PC4? or else PC1~ PC2 PC3 PC4 Eigenvector Band 1 Band 2 Band 3 Band 4 Band 1 0.587621 0.598896 0.543896 0.014209 Band 2 0.154339 0.068470 -0.217023 -0.961453 Band 3 0.294615 0.438520 -0.807980 0.260903 Band 4 0.737619 -0.666585 -0.065164 0.085646 Also do the eigen vector matrix elements correspond to factor loadings?

    MariM



    Veteran Member


    Posts:2396
    Veteran Member


    --
    10 Oct 2016 08:18 AM
    This article may help: http://www.harrisgeospati...cleID/2765/2765.aspx I don't know what ERDAS does but it may be a difference of computing the PCs on the correlatin vs covariance matrix. The default in ENVI is to use the covariance matrix.

    Bisma Qazi



    New Member


    Posts:16
    New Member


    --
    11 Oct 2016 12:34 AM
    I followed the link and it is mentioned there that 'eigenvectors is a 2D array containing the eigenvectors in its columns'. Would it be right to say then PCs are represented by columns as illustrated earlier in second matrix ?
    You are not authorized to post a reply.