+ Reply to Thread
Results 1 to 2 of 2

Thread: Principal Component Analysis - global R2

  1. #1
    Points: 8, Level: 1
    Level completed: 15%, Points required for next Level: 42

    Posts
    2
    Thanks
    0
    Thanked 0 Times in 0 Posts

    Principal Component Analysis - global R2




    Hi,

    I've extracted the principal components from a set of k studentized variables. I can compute the cumulative R2.i.e the percentage of variation explained by the first r principal components by taking the ratio of the sum of the first r eigenvalues (i.e the largest r eigenvalues) over the sum of all eigenvalues.

    The problem comes in when I try to compute the average R2. To obtain this I compute the R2 from regressions of all k original variables on the first r principal components. I then average across all k variables. Now, theoretically this should give me the same "cumulative" R2 as the one described above. I however get very different results.

    Am I missing something? I have been checking my code, and running it in parts with a very small set of variables and the two ways of calculating the cumulative variation explained do not yield the same results. Have I made a mistake thinking I should get hte same results?


    Thank you very much,
    vm

  2. #2
    Points: 8, Level: 1
    Level completed: 15%, Points required for next Level: 42

    Posts
    2
    Thanks
    0
    Thanked 0 Times in 0 Posts

    Re: Principal Component Analysis - global R2


    Nevermind. It as a code mistake after all. I was using an outdated MATLAB function which messed things up.

    Maths does work after all!

+ Reply to Thread

           




Tags for this Thread

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts






Advertise on Talk Stats