ATTRIBUTES Subcommand (PROXMAP command)

[/ATTRIBUTES [VARIABLES = varlist]
    [FUNCTION = {LINEAR }]
               {SPLINE** [DEGREE = {2*}] }
                                   {n }
                         [INKNOTS = {2*}]
                                    {n }
                         [PLACEMENT = {INTERVALS* }]
                                      {PERCENTILES}
                         [MONOTONIC = {YES*}]
                                      {NO  }
               {ORDINAL [APPROACH = {KEEPTIES}] }
                                    {UNTIE   }
               {NOMINAL}
[/ATTRIBUTES ...]

The ATTRIBUTES subcommand restricts the configuration in the common space to be a linear combination of selected attribute variables. This subcommand is used when external variables (attributes) are believed to explain the configuration, and you want to fit the PROXMAP solution within the subspace defined by these attributes. This subcommand might be repeated, allowing different sets of attributes to be transformed with different specifications.

VARIABLES

Variables specifies a list of attribute variables to be used in the analysis. All listed variables must be present in the active dataset.

The number of attributes must exceed the number of dimensions.
The number of attributes should be substantially smaller than the number of objects.
To ensure stable optimization, attribute variables should not be too highly correlated.

FUNCTION

Specifies the type of transformation to apply to the attribute variables. It includes the following options.

Option	Description
`LINEAR`	Applies a linear transformation. The transformed values remain proportional to the original attribute values.
`SPLINE`	This is the default option. Applies a smooth piecewise polynomial transformation. Additionally, spline transformation includes the following options: `DEGREE = n` indicates the degree of the polynomial pieces. The value must be an integer greater than 0. The default vaue is 2. `INKNOTS = n` indicates the number of interior knots. The value must be an integer in the interval [0, n], where `n` is the number of distinct attribute values minus two. If `INKNOTS` is greater than this number, it will be set to this maximum. The default value is 2. The maximum value for the sum of DEGREE and INKNOTS is the number of distinct attribute values minus 1. If this maximum is exceeded, the DEGREE specification is adapted. `PLACEMENT = INTERVAL` indicates that knots are placed at evenly spaced intervals over the attribute value range. `PLACEMENT = PERCENTILES` indicates that knots are placed at percentile points based on the distribution of attribute values. `MONOTONIC = YES` is set as default. The transformation must be monotonic. `MONOTONIC = NO` indicates that monotonicity is not required.
`ORDINAL`	Applies an ordinal transformation that preserves only the rank order of the attribute values. For tied values, the following options are available: `APPROACH = UNTIE` is the default setting. With this parameter, tied values become separated during transformation. `APPROACH = KEEPTIES` indicates that ties in the original data are retained in the transformation.
`NOMINAL`	Treats attribute values as unordered categories. The categorical values are replaced with numeric values in a one-to-one fashion. Only grouping information (not order) is preserved. The transformed variable can be represented in the same way as variables transformed by other transformation functions, which is by one set of transformed values (in contrast with nominal variables in the dataset or nominal property variables).

Example


GET FILE’kinship.sav’.
PROXMAP
 /SOURCEID DATA = aunt brother cousin daughter father granddaughter grandfather grandmother grandson
                  mother nephew niece sister son uncle
 /DATA PROXIMITIES
 /ATTRIBUTES VARIABLES = degree generation FUNCTION = ORDINAL
 /ATTRIBUTES VARIABLES = gender FUNCTION = NOMINAL.

In this example, proximities between kinship terms (variables aunt through uncle) are analyzed. Because these variables represent proximities, the DATA subcommand is required.
Three attribute variables are specified: degree, generation, and gender.
The attributes degree and generation are transformed ordinally, using the default setting APPROACH = UNTIE, which allows tied values to be untied.
The attribute gender is transformed nominally.