>  Docs Center  >  Libraries  >  ASTROLIB  >  PCA
Libraries

PCA

PCA

Name


    PCA

Purpose


    Carry out a Principal Components Analysis (Karhunen-Loeve Transform)

Explanation


    Results can be directed to the screen, a file, or output variables
    See notes below for comparison with the intrinsic IDL function PCOMP.

Calling Sequence


    PCA, data, eigenval, eigenvect, percentages, proj_obj, proj_atr,
            [MATRIX =, TEXTOUT = ,/COVARIANCE, /SSQ, /SILENT ]

Input Parameters


    data - 2-d data matrix, data(i,j) contains the jth attribute value
              for the ith object in the sample. If N_OBJ is the total
              number of objects (rows) in the sample, and N_ATTRIB is the
              total number of attributes (columns) then data should be
              dimensioned N_OBJ x N_ATTRIB.

Optional Input Keyword Parameters


    /COVARIANCE - if this keyword is set, then the PCA will be carried out
              on the covariance matrix (rare), the default is to use the
              correlation matrix
    /SILENT - If this keyword is set, then no output is printed
    /SSQ - if this keyword is set, then the PCA will be carried out on
              on the sums-of-squares & cross-products matrix (rare)
    TEXTOUT - Controls print output device, defaults to !TEXTOUT
              textout=1 TERMINAL using /more option
              textout=2 TERMINAL without /more option
              textout=3 <program>.prt
              textout=4 laser.tmp
              textout=5 user must open file
              textout = filename (default extension of .prt)

Optional Output Parameters


    eigenval - N_ATTRIB element vector containing the sorted eigenvalues
    eigenvect - N_ATRRIB x N_ATTRIB matrix containing the corresponding
              eigenvectors
    percentages - N_ATTRIB element containing the cumulative percentage
            variances associated with the principal components
    proj_obj - N_OBJ by N_ATTRIB matrix containing the projections of the
            objects on the principal components
    proj_atr - N_ATTRIB by N_ATTRIB matrix containing the projections of
              the attributes on the principal components

Optional Output Parameter


      MATRIX = analysed matrix, either the covariance matrix if /COVARIANCE
              is set, the "sum of squares and cross-products" matrix if
              /SSQ is set, or the (by default) correlation matrix. Matrix
              will have dimensions N_ATTRIB x N_ATTRIB

Notes


      This procedure performs Principal Components Analysis (Karhunen-Loeve
      Transform) according to the method described in "Multivariate Data
      Analysis" by Murtagh & Heck [Reidel : Dordrecht 1987], pp. 33-48.
      See http://astro.u-strasbg.fr/~fmurtagh/mda-sw/
      Keywords /COVARIANCE and /SSQ are mutually exclusive.
      The printout contains only (at most) the first seven principle
      eigenvectors. However, the output variables EIGENVECT contain
      all the eigenvectors
     
      Different authors scale the covariance matrix in different ways.
      The eigenvalues output by PCA may have to be scaled by 1/N_OBJ or
      1/(N_OBJ-1) to agree with other calculations when /COVAR is set.
      PCA uses the non-standard system variables !TEXTOUT and !TEXTUNIT.
      These can be added to one's session using the procedure ASTROLIB.
      The intrinsic IDL function PCOMP duplicates most
      most of the functionality of PCA, but uses different conventions and
      normalizations. Note the following:
  (1) PCOMP requires a N_ATTRIB x N_OBJ input array; this is the transpose
        of what PCA expects
  (2) PCA uses standardized variables for the correlation matrix: the input
        vectors are set to a mean of zero and variance of one and divided by
        sqrt(n); use the /STANDARDIZE keyword to PCOMP for a direct comparison.
  (3) PCA (unlike PCOMP) normalizes the eigenvectors by the square root
        of the eigenvalues.
  (4) PCA returns cumulative percentages; the VARIANCES keyword of PCOMP
        returns the variance in each variable
  (5) PCOMP divides the eigenvalues by (1/N_OBJ-1) when the covariance matrix
          is used.

Example


      Perform a PCA analysis on the covariance matrix of a data matrix, DATA,
      and write the results to a file
      IDL> PCA, data, /COVAR, t = 'pca.dat'
      Perform a PCA analysis on the correlation matrix. Suppress all
      printing, and save the eigenvectors and eigenvalues in output variables
      IDL> PCA, data, eigenval, eigenvect, /SILENT

Procedures Called


      TEXTOPEN, TEXTCLOSE

Revision History


      Immanuel Freedman (after Murtagh F. and Heck A.). December 1993
      Wayne Landsman, modified I/O December 1993
      Fix MATRIX output, remove GOTO statements W. Landsman August 1998
      Changed some index variable to type LONG W. Landsman March 2000
      Fix error in computation of proj_atr, see Jan 1990 fix in
      http://astro.u-strasbg.fr/~fmurtagh/mda-sw/pca.f W. Landsman Feb 2008



© 2024 NV5 Geospatial Solutions, Inc. |  Legal
   Contact Us