PLOT Subcommand (PROXMAP command)

[/PLOT [DATA [([HISTOGRAM]
       [WEIGHTS [([HISTOGRAM]
       [INITIAL [([CONFIGURATION]
                  [MULTIVARIATEBIPLOT]
                  [PROPERTIES]
                  [SCREE]
                  [STRESS])]]
       [MODEL [([CONFIGURATION**]
                [MULTIVARIATEBIPLOT]
                [REDUCEDRANKBIPLOT]
                [TREE]
                [NEIGHBORS[({2*})]]
                            {n }
                [THRESHOLD[({0.33*})]]
                             {value}
                [INDIVIDUAL[(srclist)]]
                [SPACEWEIGHTS]
                [TRANSFORMATION[(srclist)]]
                [RESIDUALS[(srclist)]]
                [FIT[(srclist)]]
                [SHEPARD[(srclist)]])]]
       [ATTRIBUTES [([BIPLOT[(varlist)]
                     [PROJECTIONS]
                     [TRANSFORMATION[(varlist)]]
                     [RESIDUALS[(varlist)]])]]
       [PROPERTIES [([BIPLOT[(varlist)]
                     [PROJECTIONS]
                     [TRANSFORMATION[(varlist)]]
                     [RESIDUALS[(varlist)]])]]
       [CRITERIA [([ITERATIONS]
                   [STRESS])]]]

PLOT controls which plots are displayed. The initial and common space (bubble plots) are always displayed. All other PLOT output is optional.

  • All main keywords (for example, CONFIGURATION, DISTANCES) can be followed by one or more sub keywords in parentheses to fine-tune the output.
  • When the main keyword is specified, at least one sub keyword should be given.
  • Inapplicable keywords are silently ignored, that is, specifying a PLOT keyword while no (applicable) corresponding subcommand is specified or implied, has no effect, without issuing a warning. For example, /PLOT MODEL (SPACEWEIGHTS) when an individual difference model is not specified.
  • Objects are labeled in plots according to the specification of the LABELS keyword in the OBJECTID subcommand.
DATA
Plots of the derived proximities.

The keyword HISTOGRAM plots a histogram per source of the derived proximities.

WEIGHTS
Displays the plots of the proximity weights.

HISTOGRAM indicates a histogram per source of the derived proximity weights.

INITIAL
Plots related to the initial configuration.
Keyword Description
CONFIGURATION A plot of the initial space.
MULTIVARIATEBIPLOT Biplot of the initial space for multivariate data, showing vectors (directions) for variables with non-Nominal Measurement level, or points (centroids) for variables with Nominal Measurement level.
PROPERTIES Biplot of the initial space that includes vectors (directions) for variables specified in the PROPERTIES subcommand, applying an initial linear transformation (regardless of any other transformation functions specified under PROPERTIES).
SCREE If the starting configuration is based on classical scaling, a scree plot (eigenvalues versus dimensionality) is available.
STRESS A bubble plot of the initial space showing circles around the object points to indicate the contribution to the Normalized Stress.
MODEL
Plots of the common space, the individual spaces, biplots, and the transformations.
Keywords for the common space:
Keyword Description
CONFIGURATION Plot of the common space.
MULTIVARIATEBIPLOT Biplot of the common space including vectors (directions) for the variables from which the proximities are derived for multivariate data, for variables with non-Nominal Measurement level, or points (centroids) for variables with Nominal Measurement level.
TREE Displays a minimum spanning tree that connects all object points using links with the smallest total distance. This plot can be used to derive a single-link hierarchical structure.
NEIGHBORS (n) Shows a nearest neighbor graph in which each object is linked to its n closest neighbors. The minimum value for n is 1; the maximum is the number of objects minus one. The default value is n = 2.
THRESHOLD (value) Displays an undirected threshold graph based on Waern links. This plot visually compares proximities with distances by connecting objects that are similar according to their proximity values. A numeric value can be specified to include only the n% of object pairs with the smallest proximities (default is 33%).
Keywords for individual spaces (only applicable when the model is not IDENTITY):
Keyword Description
INDIVIDUAL (srclist) Plots of individual source spaces for the sources listed in srclist. For the Generalized Euclidean model with reduced rank 1, the individual (1-dimensional) spaces are not shown. These 1-dimensional spaces are depicted as vectors in the common space when the keyword REDUCEDRANKBIPLOT is used.
SPACEWEIGHTS Plot of the dimension weights (for the generalized Euclidean model, the dimension component of the space weights).

For the weighted identity (dilation) model, the weights are 1-dimensional and are shown in a plot with the source names or labels on the horizontal axis and the weights on the vertical axis.

For the weighted Euclidean model, the weights are multidimensional and are plotted in a plot with a dimension on each axis. These weights are not plotted for the Generalized Euclidean model because here the weights refer to different, rotated, individual spaces.

REDUCEDRANKBIPLOT Biplot of the common space and the dimension component of the space weights for a rank-1 model. Vectors in the common space have the same length.
Keywords for transformations, residuals, fit and Shepard plots (a single plot is shown for unconditional models; for MATRIX conditional models, a separate plot is shown for each source in srclist):
Keyword Description
TRANSFORMATION (srclist) Plots of transformed proximities (vertical axis) versus original proximities (horizontal axis).
RESIDUALS (srclist) Plots of residuals, defined as the differences between the transformed proximities and the distances in the configuration (vertical axis), plotted against the transformed proximities (horizontal axis).
FIT (srclist) Plots of distances in the configuration (vertical axis) versus transformed proximities (horizontal axis).
SHEPARD (srclist) Plots of transformed proximities and corresponding distances (vertical axis) versus original proximities (horizontal axis).
ATTRIBUTES
Plots involving attributes. See the following keywords for attribute variables (as specified on the ATTRIBUTES subcommand):
Keyword Description
BIPLOT (varlist) Displays the restricted space including vectors (directions) for transformed attributes, including nominally transformed attributes.
PROJECTIONS Shows the projections of the object points onto the attribute vectors from the biplot. The plot uses attribute names on the horizontal axis and the projected values (fitted values) on the vertical axis. This plot complements the biplot by showing projected points explicitly, which are omitted from the main biplot to avoid visual clutter.
TRANSFORMATION (varlist) Displays a transformation plot for each attribute that is listed in varlist.
RESIDUALS (varlist) Displays a residual plot for each attribute in varlist.
PROPERTIES
Plots involving properties. See the following keywords for property variables (as specified on the PROPERTIES subcommand):
Keyword Description
BIPLOT (varlist) Displays the configuration that includes vectors (lines) for the specified property variables.
TRANSFORMATION (varlist) Shows a transformation plot for each property in varlist.
PROJECTIONS Displays the projections of the object points onto the property vectors from the biplot. The property names appear on the horizontal axis, and the projected values are shown on the vertical axis. This plot supplements the biplot by explicitly showing the projected points, which are omitted in the main biplot to reduce label clutter. Each property is represented as a vertical line with the projected values indicated along it.
RESIDUALS (varlist) Displays a residual plot for each property in varlist.
CRITERIA
Plots related to convergence and dimensionality criteria.

ITERATIONS shows the iteration history of the algorithm. The plot displays Normalized Stress values on the vertical axis and the iteration number on the horizontal axis.

STRESS shows the bubble plot of the common space, with circles around object points. The size of each circle indicates the relative contribution of the object to the overall Normalized Stress.

Example

GET FILE ’States.sav’.
  PROXMAP
  /SOURCEID NAME=$SRC001 LABEL='popul' DATA=popul
  /SOURCEID NAME=$SRC002 LABEL='income' DATA=income
  /SOURCEID NAME=$SRC003 LABEL='illit' DATA=illit
  /SOURCEID NAME=$SRC004 LABEL='life' DATA=life
  /SOURCEID NAME=$SRC005 LABEL='homic' DATA=homic
  /SOURCEID NAME=$SRC006 LABEL='school' DATA=school
  /SOURCEID NAME=$SRC007 LABEL='freeze' DATA=freeze
  /SOURCEID NAME=$SRC008 LABEL='fail' DATA=fail
  /DATA MULTIVARIATE=STANDARDIZED
  /OBJECTID LABELS=STATE
  /MODEL GENERALIZED RANK = 1
  /PROPERTIES VARIABLES=income illit life homic school fail FUNCTION=LINEAR
  /PRINT INITIAL (DIAGNOSTICS) MODEL (SPACEWEIGHTS) ATTRIBUTES (TRANSFORMED FITTED DIRECTIONS)
    PROPERTIES (TRANSFORMED FITTED DIRECTIONS) CRITERIA (SUMMARY DIAGNOSTICS SCREE ITERATIONS STRESS)
  /PLOT INITIAL (CONFIGURATION ) MODEL (REDUCEDRANKBIPLOT TREE) PROPERTIES (PROJECTIONS) CRITERIA   
(ITERATIONS).
  • In this example, each variable defines a source. The Generalized Euclidean model is fitted, with two dimensions in the common space.
  • Each individual space has rank 1 because each source contains only one variable. With this model, each source (variable) is represented as a subspace of the common space.
  • A multivariate biplot of the common space is obtained with the original variables in the dataset, a minimum spanning tree in the common space,
  • Projections of the object points onto the property vectors in the common space
  • The reduced rank biplot, where the direction coordinates are obtained directly from the spaceweights
  • A plot is added that depicts a summary of the iteration history.