CRITERIA Subcommand (FACTOR command)

CRITERIA controls extraction and rotation criteria.

  • CRITERIA can be specified anywhere after VARIABLES and MISSING.
  • Only explicitly specified criteria are changed. Unspecified criteria keep their defaults.
  • Multiple CRITERIA subcommands are allowed. Changes made by a previous CRITERIA subcommand are overwritten by a later CRITERIA subcommand.
  • Any CRITERIA subcommands specified after the last EXTRACTION subcommand have no effect on extraction.
  • Any CRITERIA subcommands specified after the last ROTATION subcommand have no effect on rotation.

The following keywords on CRITERIA apply to extractions:

FACTORS(n). Number of factors extracted. The default is the number of eigenvalues greater than MINEIGEN. When specified, FACTORS overrides MINEIGEN.

MINEIGEN(n). Minimum eigenvalue used to control the number of factors extracted. If METHOD=CORRELATION, the default is 1. If METHOD=COVARIANCE, the default is computed as (Total Variance/Number of Variables)*n, where Total Variance is the total weighted variance principal components or principal axis factoring extraction and the total image variance for image factoring extraction.

ECONVERGE(n). Convergence criterion for extraction. The default is 0.001.

The following keywords on CRITERIA apply to rotations:

RCONVERGE(n). Convergence criterion for rotation. The default is 0.0001.

KAISER. Kaiser normalization in the rotation phase. This is the default. The alternative is NOKAISER.

NOKAISER. No Kaiser normalization.

The following keywords on CRITERIA apply to both extractions and rotations:

ITERATE(n). Maximum number of iterations for solutions in the extraction or rotation phases. The default is 25.

DEFAULT. Reestablish default values for all criteria.

Example

FACTOR VARIABLES=V1 TO V12
  /CRITERIA=FACTORS(6) 
  /EXTRACTION=PC
  /ROTATION=NOROTATE
  /PLOT=ROTATION.
  • This example analyzes all variables between and including V1 and V12 in the active dataset.
  • Six factors are extracted using the default principal components method, and the factor pattern matrix is not rotated.
  • PLOT sends all extracted factors to the graphics editor and shows a 3-D plot of the first three factors.