Google
 

Trailing-Edge - PDP-10 Archives - decuslib20-02 - decus/20-0026/varmx.doc
There are 2 other files named varmx.doc in the archive. Click here to see a list.
SUBROUTINE VARMX

PURPOSE
   PERFORM ORTHOGONAL ROTATIONS OF A FACTOR MATRIX.  THIS
   SUBROUTINE NORMALLY OCCURS IN A SEQUENCE OF CALLS TO SUB-
   ROUTINES CORRE, EIGEN, TRACE, LOAD, VARMX IN THE PERFORMANCE
   OF A FACTOR ANALYSIS.

USAGE
   CALL VARMX (M,K,A,NC,TV,H,F,D,IER)

DESCRIPTION OF PARAMETERS
   M	 - NUMBER OF VARIABLES AND NUMBER OF ROWS OF MATRIX A.
   K	 - NUMBER OF FACTORS.
   A	 - INPUT IS THE ORIGINAL FACTOR MATRIX, AND OUTPUT IS
	   THE ROTATED FACTOR MATRIX.  THE ORDER OF MATRIX A
	   IS M X K.
   NC	 - OUTPUT VARIABLE CONTAINING THE NUMBER OF ITERATION
	   CYCLES PERFORMED.
   TV	 - OUTPUT VECTOR CONTAINING THE VARIANCE OF THE FACTOR
	   MATRIX FOR EACH ITERATION CYCLE.  THE VARIANCE PRIOR
	   TO THE FIRST ITERATION CYCLE IS ALSO CALCULATED.
	   THIS MEANS THAT NC+1 VARIANCES ARE STORED IN VECTOR
	   TV.	MAXIMUM NUMBER OF ITERATION CYCLES ALLOWED IN
	   THIS SUBROUTINE IS 50.  THEREFORE, THE LENGTH OF
	   VECTOR TV IS 51.
   H	 - OUTPUT VECTOR OF LENGTH M CONTAINING THE ORIGINAL
	   COMMUNALITIES.
   F	 - OUTPUT VECTOR OF LENGTH M CONTAINING THE FINAL
	   COMMUNALITIES.
   D	 - OUTPUT VECTOR OF LENGTH M CONTAINING THE DIFFERENCES
	   BETWEEN THE ORIGINAL AND FINAL COMMUNALITIES.
   IER	 - ERROR INDICATOR
	   IER=0 - NO ERROR
	   IER=1 - CONVERGENCE WAS NOT ACHIEVED IN 50 CYCLES
		   OF ROTATION

REMARKS
   IF VARIANCE COMPUTED AFTER EACH ITERATION CYCLE DOES NOT
   INCREASE FOR FOUR SUCCESSIVE TIMES, THE SUBROUTINE STOPS
   ROTATION.

SUBROUTINES AND FUNCTION SUBPROGRAMS REQUIRED
   NONE

METHOD
   KAISER'S VARIMAX ROTATION AS DESCRIBED IN 'COMPUTER PROGRAM
   FOR VARIMAX ROTATION IN FACTOR ANALYSIS' BY THE SAME AUTHOR,
   EDUCATIONAL AND PSYCHOLOGICAL MEASUREMENT, VOL XIX, NO. 3,
   1959.