X
PrevPrev Go to previous topic
NextNext Go to next topic
Last Post 29 May 2009 07:35 AM by  anon
PCOMP - eigenql
 3 Replies
Sort:
You are not authorized to post a reply.
Author Messages

anon



New Member


Posts:
New Member


--
29 May 2009 07:35 AM
    Hello.   I'm attempting a large p.c.a. with a high number of variables. pcomp and eigenql keep bugging out when I supply the matrix, but works fine when I supply only a subsection. IDL> .r run_eof_test % Compiled module: $MAIN$. MULTIPLOT: device=X ncol=       2 nrow=       2 charsize=      1.25000 % TRIQL: Too many iterations in tqli. Obviously there is a limit to the size of the number of variables. Any way to over-ride this? Is it dependent on hardware? or the idl routine? Appreciate any help. Thanks. Craig J Wallace  

    Deleted User



    New Member


    Posts:
    New Member


    --
    10 Jun 2009 09:22 AM
    Hi Brandon Thanks for the reply. I'm on a 64-bit Linux box with 8gb RAM.... the array of data supplied to PCOMP is [1113, 73], and the correlation matrix calculated in the first step of PCOMP will be [1113,1113]. Best regards Craig.

    MariM



    Veteran Member


    Posts:2396
    Veteran Member


    --
    10 Jun 2009 09:51 AM
    The TQLI error is caused by the routine encountering a singularity in the covariance matrix.  Below is a tech tip describing why you can get this error running a PCA in ENVI but ENVI uses the underlying IDL routine PCOMP, so the information should be helpful to track down the source of the error in your data: http://www.ittvis.com/ser...echtip.asp?ttid=3343

    Deleted User



    New Member


    Posts:
    New Member


    --
    12 Jun 2009 04:41 AM
    Thanks.   Craig.
    You are not authorized to post a reply.