CORAN85.DOC The new (May 1985) Correspondence Analysis program (CORAN85 in the following) no longer requires transposition of the image set. A sequential file is created, the records of which contain the NPIX pixels passed by the mask. One record is created for each image, in the order the images are entered. Two options are available: In-Core computation and Stochastic approximation. When the in-core option is selected, the entire matrix of dimension NPIX x NPIX is stored in core. Because of this, it should not be used for projects involving more than 200 pixels in the mask. The method of stochastic approximation, developed by Lebart specifically to accomodate large data sets, works by going through the sequential file in seven passes, taking random samples of the pixels. CORAN85 is implemented, in modular form, as part of SPIDER. At present, there are eight operations: CA_SI Initializes CORAN by preparing a sequential file. No background correction used. CA_SIB as 'CA_SI', but with background correction. CA_S Runs CORAN, using sequential file created by either 'CA_SI' or 'CA_SIB', and puts the results into image-coordinate file, pix- coordinate file, and eigenvector file. CA_SM prints out selected factor maps, either of image coordinates or of pixel coordinates. CA_SME types out list of eigenvalues in addition to printing the map. (If only eigenvalues are desired, answer "(P)ixel or (I)mage coor- dinates" with any symbol other than 'P' or 'I'.) CA_SR reconstitutes selected images using a subset of the eigenvectors. CA_SRD produces differential reconstituted images (i.e., the O-factor image has been subtracted). CA_SRI produces importance images (i.e., differential reconstituted images without weighting of the pixels). The auxiliary files containing image coordinates, pixel coordinates, and eigenvectors are sequential. All coordinates are stored as floating point numbers. Subroutines called by UTIL1 Directory OLB ====================================== ========== ====== JPMSK1.FOR (CA_S, CA_SI, CA_SIB) .FORTRAN] REAL SGRAF.FOR (CA_SM, CA_SME) .FORTRAN] JP JPMSK2.FOR (CA_SR, CA_SRD, CA_SRI) .FORTRAN] REAL Subroutine Called by Stored in Purpose ============ =========== =========== =========================== BORNS HPLAN HPLAN Finds extreme values of array CPROJ STOCHA STOC DOGRAF EPUR4 HPLAN HPLAN Collapses map according to S.D. multiple GETCOO SCOR STOC Computes coos and distances for rows and columns GETSPE GSMOD ITPOW STOC Gram-Schmidt orthonormal- ization HPLAN SGRAF1 GRAPH IMSEQ4 JPMSK2 JPMSK2 Selects images INCORE STOC Driver of CA in-core ITPOW STOCHA STOC Iterated power method MSKDEC JPMSK2 Decodes mask array RECDUM JPMSK2 RECDUM Reconstitutes dummy images using factor coordinates supplied RECONS JPMSK2 RECONS Reconstitute selected image set REW REW Rewinds and skips header SCOR SCORAN STOC Driver of CA/analysis SCORAN SCORAN Driver of CA/memory management SEN3A STOC Uniform pseudo random generator SGRAF SGRAF Driver of graphic output SGRAF1 SGRAF SGRAF Graphic output organizer SRIP2 SRIPE SRIPE SRIPE JPMSK2 SRIPE Initialize RCONS, perform reconstitution STOCHA SCOR STOC Driver of CA/analysis/ stochastic TRIDI STOC VPROP / CPROJ STOC \ INCORE JPMSK1 JPMSK1 Driver of sequential file preparation from SPIDER files and CORAN analysis JPMSK2 JPMSK2 Driver of reconstitution of SPIDER images Page 3 List of source files: Name Compile Library =========== ============ ================ HPLAN I4 [SCOR]JP JPMSK1 I2 [SPIDER]REAL JPMSK2 I2 REAL RECONS I4 JP RECDUM I4 JP REW I4 JP SGRAF I2 REAL SRIPE I4 JP STOC I4 JP