PLOT Subcommand (CATPCA command)

The PLOT subcommand controls the display of plots. The default keywords are OBJECT and LOADING. That is, the two keywords are in effect when the PLOT subcommand is omitted or when the PLOT subcommand is given without any keyword. If a keyword is duplicated (for example, /PLOT = RESID RESID), only the last keyword is effective. If the keyword NONE is used with other keywords (for example, /PLOT = RESID NONE LOADING), only the keywords following NONE are effective. When keywords contradict, the later keyword overwrites the earlier keywords.

  • All the variables to be plotted must be specified on the ANALYSIS subcommand.
  • If the variable list following the keywords CATEGORIES, TRANS, RESID, and PROJCENTR is empty, it will cause a syntax error, and the procedure will terminate.
  • The variables in the variable list for labeling the object point following OBJECT, BIPLOT, and TRIPLOT must be specified on the VARIABLES subcommand but need not appear on the ANALYSIS subcommand. This flexibility means that variables that are not included in the analysis can still be used to label plots.
  • The keyword TO in a variable list can only be used with variables that are in the ANALYSIS subcommand, and TO applies only to the order of the variables in the ANALYSIS subcommand. For variables that are in the VARIABLES subcommand but not in the ANALYSIS subcommand, the keyword TO cannot be used. For example, if /VARIABLES = v1 TO v5 and /ANALYSIS = v2 v1 v4, then /PLOT OBJECT(v1 TO v4) will give two object plots, one plot labeled with v1 and one plot labeled with v4.
  • For multidimensional plots, all of the dimensions in the solution are produced in a matrix scatterplot if the number of dimensions in the solution is greater than 2 and the NDIM plot keyword is not specified; if the number of dimensions in the solution is 2, a scatterplot is produced.

The following keywords can be specified:

OBJECT(varlist)(n). Plots of the object points. Following the keyword, a list of variables in parentheses can be given to indicate that plots of object points labeled with the categories of the variables should be produced (one plot for each variable). The variables to label the objects must be specified on the VARIABLES subcommand but need not appear on the ANALYSIS subcommand. If the variable list is omitted, a plot that is labeled with case numbers is produced.

CATEGORY(varlist)(n). Plots of the category points. Both the centroid coordinates and the vector coordinates are plotted. A list of variables must be given in parentheses following the keyword. For variables with optimal scaling level MNOM, categories are in the centroids of the objects in the particular categories. For all other optimal scaling levels, categories are on a vector through the origin.

LOADING(varlist (CENTR(varlist)))(l). Plot of the component loadings optionally with centroids. By default, all variables with an optimal scaling level that results in vector quantification (that is, SPORD, SPNOM, ORDI, NOMI, and NUME) are included in this plot. LOADING can be followed by a varlist to select the loadings to include in the plot. When "LOADING(" or the varlist following "LOADING(" is followed by the keyword CENTR in parentheses, centroids are included in the plot for all variables with optimal scaling level MNOM. CENTR can be followed by a varlist in parentheses to select MNOM variables whose centroids are to be included in the plot. When all variables have the MNOM scaling level, this plot cannot be produced.

TRANS(varlist(n))(n). Transformation plots per variable (optimal category quantifications against category indicators). Following the keyword, a list of variables in parentheses must be given. MNOM variables in the varlist can be followed by a number of dimensions in parentheses to indicate that you want to display p transformation plots, one plot for each of the first p dimensions. If the number of dimensions is not specified, a plot for the first dimension is produced.

RESID(varlist(n))(n). Plot of residuals per variable (approximation against optimal category quantifications). Following the keyword, a list of variables in parentheses must be given. MNOM variables in the varlist can be followed by a number of dimensions in parentheses to indicate that you want to display p residual plots, one plot for each of the first p dimensions. If the number of dimensions is not specified, a plot for the first dimension is produced.

BIPLOT(keyword(varlist)) (varlist)(n). Plot of objects and variables. The coordinates for the variables can be chosen to be component loading or centroids, using the LOADING or CENTR keyword in parentheses following BIPLOT. When no keyword is given, component loadings are plotted. When NORMALIZATION = INDEPENDENT, this plot is incorrect and therefore not available. Following LOADING or CENTR, a list of variables in parentheses can be given to indicate the variables to be included in the plot. If the variable list is omitted, a plot including all variables is produced. Following BIPLOT, a list of variables in parentheses can be given to indicate that plots with objects that are labeled with the categories of the variables should be produced (one plot for each variable). The variables to label the objects must be specified on the VARIABLES subcommand but need not appear on the ANALYSIS subcommand. If the variable list is omitted, a plot with objects labeled with case numbers is produced.

TRIPLOT(varlist(varlist))(n). A plot of object points, component loadings for variables with an optimal scaling level that results in vector quantification (that is, SPORD, SPNOM, ORDI, NOMI, and NUME ), and centroids for variables with optimal scaling level MNOM. Following the keyword, a list of variables in parentheses can be given to indicate the variables to include in the plot. If the variable list is omitted, all variables are included. The varlist can contain a second varlist in parentheses to indicate that triplots with objects labeled with the categories of the variables in this variable list should be produced (one plot for each variable). The variables to label the objects must be specified on the VARIABLES subcommand but need not appear on the ANALYSIS subcommand. If this second variable list is omitted, a plot with objects labeled with case numbers is produced. When NORMALIZATION = INDEPENDENT, this plot is incorrect and therefore not available.

JOINTCAT(varlist)(n). Joint plot of the category points for the variables in the varlist. If no varlist is given, the category points for all variables are displayed.

PROJCENTR(varname, varlist)(n). Plot of the centroids of a variable projected on each of the variables in the varlist. You cannot project centroids of a variable on variables with MNOM optimal scaling level; thus, a variable that has MNOM optimal scaling level can be specified as the variable to be projected but not in the list of variables to be projected on. When this plot is requested, a table with the coordinates of the projected centroids is also displayed.

VAF. Barcharts of variable variance accounted for. There is one barchart for each dimension and one for the total variance accounted for over all dimensions..

LDELLAREA(threshold). Confidence ellipses for loading plots. If RESAMPLE=BOOTSTRAP and DIMENSION=2, confidence ellipses are plotted. You can control the display of loadings with their confidence ellipses in the plot by specifying a threshold ellipse area in parentheses of the general form: (GT|LT STDEV|AREA value). STDEV represents the mean area plus the number of standard deviations specified for the value. You can use the greater than (>) and less than signs (<) instead of GT and LT. The default setting is (> AREA 0). This displays all loadings with confidence ellipses.

OBELLAREA(threshold). Confidence ellipses for object plots. If RESAMPLE=BOOTSTRAP and DIMENSION=2, confidence ellipses are plotted. You can control the display of objects with their confidence ellipses in the plot by specifying a threshold ellipse area in parentheses of the general form: (GT|LT STDEV|AREA value). STDEV represents the mean area plus the number of standard deviations specified for the value. You can use the greater than (>) and less than signs (<) instead of GT and LT. The default setting is (> STDEV 2). This displays all objects with confidence ellipses.

CTELLAREA(threshold). Confidence ellipses for category plots. If RESAMPLE=BOOTSTRAP and DIMENSION=2, confidence ellipses are plotted. You can control the display of loadings with their confidence ellipses in the plot by specifying a threshold ellipse area in parentheses of the general form: (GT|LT STDEV|AREA value). STDEV represents the mean area plus the number of standard deviations specified for the value. You can use the greater than (>) and less than signs (<) instead of GT and LT. The default setting is (> AREA 2). This displays all categories with confidence ellipses.

NELLPNT(integer). Number of ellipse contour points. If RESAMPLE=BOOTSTRAP and DIMENSION=2, confidence ellipses are plotted as a path between a number of points on the ellipse contours. The number of these points influences how smooth the ellipses look. The default number of ellipse contour points is 40.

NONE. No plots.

  • For all keywords that allow a variable list, the user can specify an optional parameter l in parentheses after the variable list in order to control the global upper boundary of variable name/label and value label lengths in the plot. Note that this boundary is applied uniformly to all variables in the list. The label length parameter l can take any non-negative integer that is less than or equal to the applicable maximum length (64 for variable names, 255 for variable labels, and 60 for value labels). If l = 0, names/values instead of variable/value labels are displayed to indicate variables/categories. If l is not specified, CATPCA assumes that each variable name/label and value label is displayed at its full length. If l is an integer that is larger than the applicable maximum, we reset it to the applicable maximum but do not issue a warning. If a positive value of l is given but some or all variables/category values do not have labels, then, for those variables/values, the names/values themselves are used as the labels.

In addition to the plot keywords, the following keyword can be specified:

NDIM(value,value). Dimension pairs to be plotted. NDIM is followed by a pair of values in parentheses. If NDIM is not specified or is specified without parameter values, a matrix scatterplot including all dimensions is produced.

  • The first value (an integer that can range from 1 to the number of dimensions in the solution minus 1) indicates the dimension that is plotted against higher dimensions.
  • The second value (an integer that can range from 2 to the number of dimensions in the solution) indicates the highest dimension to be used in plotting the dimension pairs.
  • The NDIM specification applies to all requested multidimensional plots.