CRITERIA Subcommand (PROXMAP command)

[/CRITERIA [DIMENSIONS = {2* }]]
                         {n  }
           [MAXIMUM = {1000* }]
                       {n    }
           [MINIMUM = {0.000001* }]
                      {value     }
           [RELATIVE = {0.000000000001* }]
                       {value           }
           [ABSOLUTE = {0.000001* }]
                       {value     }
           [STANDARDIZATION = {"N"* }]
                              {"N-1"}
           [REFLECTION = {NONE*                            }]
                         {CORRELATION                      }
                         {QUADRANT                         }
                         {DIMENSIONS NUMBERS = = {+1} [...]}
                                                 {-1       }

The CRITERIA subcommand specifies convergence control settings for the PROXMAP algorithm. These criteria determine when the iterative algorithm should stop, either because a solution has been reached within a specified tolerance or because a maximum number of iterations has been performed.

The CRITERIA subcommand is also used to do the following actions:

• Set the dimensionality of the solution.
• Specify termination conditions for the iterative optimization process.
• Define standardization rules and reflection behavior for output dimensions.

DIMENSIONS = n

Specifies the number of dimensions in the solution. The value must be greater than or equal to 2 and less than or equal to the number of objects − 1. The default value is 2.

MAXIMUM = n

Specifies the maximum number of iterations allowed. The default value is 1000. Reducing this number may lead to faster results but less precise solutions, since the (local) minimum of the loss function has not been reached yet.

MINIMUM = value

Minimum normalized stress value required for convergence. The algorithm stops when normalized stress falls below this threshold.

Range: [0, 1], Default: 0.000001

RELATIVE = value

Relative change in stress required between iterations. The algorithm stops if the change in normalized stress between iterations is smaller than this value.

Range: [0, 1]. Default: 1.0e−12

ABSOLUTE = value

Absolute configuration difference threshold. Non-negative value. The default value is 1.0e−6. The algorithm stops when the Euclidean norm of the differences between successive configurations is less than or equal to this value. This is the most stringent convergence criterion and helps prevent premature termination by considering even small point movements.

STANDARDIZATION = 'N' or 'N-1'

This keyword specifies how the variables in multivariate data are standardized before proximities are derived and how attribute and property variables are scaled.

N standardizes by the number of cases (default behavior).

N-1 standardizes by degrees of freedom (that is, one less than the number of cases).

Note: The keyword must be enclosed in quotes, as in 'N' or 'N-1'.

REFLECTION

Determines how the output space is reflected for interpretability:

Option	Description
NONE	No reflection is applied
CORRELATION	Reflects dimensions so that coordinate values are positively correlated with the input data.
QUADRANT	Reflects dimensions such that the maximum values lie in the positive quadrant.
DIMENSION	(±1 ...) Allows user-defined reflection per dimension. Each value must be either +1 (no reflection) or −1 (reflect). The number of values must match the number of dimensions.

PROXMAP
GET FILE ’Ekman.sav’.
PROXMAP /SOURCEID DATA = nm434 nm445 nm465 nm472 nm490 nm504 nm537 nm555 nm584 nm600 nm610 nm628 nm651
                         nm674
 /DATA = PROXIMITY
 /CRITERIA DIMENSIONS = 2
                 MAXIMUM = 200
                 RELATIVE = 0.00001
                 REFLECTION = DIMENSIONS NUMBERS = 1 -1.