Introduction to Classification Panels :: SAS(R) 9.3 Graph Template Language: User's Guide

Classification Panels in the GTL

A classification panel is a graph with one or more cells in which each cell shows a common graph (called a prototype). The prototypes that are displayed in the cells result from dividing input data into subsets that are determined by the values of one or more classification variables. GTL provides two layouts that can produce classification panels:

LAYOUT DATAPANEL

supports a list of class variables. The number of rows and columns are controlled by statement options. Each cell is labeled with the class variable values in the cell header.

LAYOUT DATALATTICE

supports up to two class variables, one for a row variable and one for a column variable. One row of cells is created for each value of the row class variable, and one column is created for each value of the column class variable. The rows and columns are labeled.

The LAYOUT DATAPANEL Statement

The example in this section uses the LAYOUT DATAPANEL statement to specify a list of two classification variables: DIVISION (two distinct values) and PRODUCT (three distinct values). Six combinations (crossings) of these unique values are possible, which produces a panel with six cells.

Notice the following details about the LAYOUT DATAPANEL statement:

The CLASSVARS= option on the LAYOUT DATAPANEL statement can specify a list of one or more classifiers.
In the resulting graph, the data crossings are identified by the cell headers.

The following template code generates Classification Panel Created with LAYOUT DATAPANEL.

proc template;
 define statgraph datapanel_intro;
  begingraph;
    entrytitle "Office Furniture Sales";
    layout datapanel classvars=(product division) / columns=2;
      layout prototype;
        seriesplot x=month y=actual;
      endlayout;
    endlayout;
  endgraph;
 end;
run;

Classification Panel Created with LAYOUT DATAPANEL

In the template code, notice the LAYOUT PROTOTYPE block, which is inside the LAYOUT DATAPANEL block. This nested block, a required part of the DATAPANEL layout, defines the graphical content of all of the cells. The COLUMNS=2 setting forces a DATAPANEL layout to display the cells in a two-column organization. The actual number of rows that are generated depends on the number of crossings that are in the data.

For some data, the number of data crossings can be quite large. Thus, when rendering the graph for a classification panel, it is common to use a WHERE expression to limit the number of crossings:

proc sgrender data=sashelp.prdsale template=datapanel_intro;
  where country="U.S.A." and region="EAST" and
        product in ("CHAIR" "DESK" "TABLE") ;
  format actual dollar.;
run;

The following schematic shows the general organization of a graph that is produced with the DATAPANEL layout. If the template code does not use the sidebar areas that are shown in the schematic, that space is reclaimed in the graph. Also, the order in which you specify the classification variables affects the cell ordering. The graph that is represented by the schematic could be produced with CLASSVARS=(classvar1 classvar2).

The LAYOUT DATALATTICE Statement

The example in this section uses the LAYOUT DATALATTICE statement to specify the same two classification variables: DIVISION and PRODUCT. Notice the following details about the LAYOUT DATALATTICE statement:

One of the ROWVAR= or COLUMNVAR= arguments is required. Both can be specified. Each specifies a single classification variable, enabling you to specify either one or two classifiers for the graph.
In the resulting graph, the data crossings are identified by row or column headers.
The default number of columns equals the number of unique values for the COLUMNVAR classifier.
The default number of rows equals the number of unique values for the ROWVAR classifier.

The following template code generates Classification Panel Created with LAYOUT DATALATTICE.

proc template;
define statgraph datalattice_intro;
  begingraph;
    entrytitle "Office Furniture Sales";
    layout datalattice rowvar=product columnvar=division;
      layout prototype;
        seriesplot x=month y=actual;
      endlayout;
    endlayout;
  endgraph;
end;
run;

proc sgrender data=sashelp.prdsale template=datalattice_intro;
  where country="U.S.A." and region="EAST" and
        product in ("CHAIR" "DESK" "TABLE");
  format actual dollar.;
run;

Classification Panel Created with LAYOUT DATALATTICE

In this example, the grid dimensions are automatically determined by the number of distinct values of the classifiers PRODUCT and DIVISION.

The following schematic shows the general organization of a graph that is produced with the DATALATTICE layout. As with a DATAPANEL layout, if the sidebar areas are not used, that space is reclaimed. Notice that the sidebar area is between the cells and the row/column headers.

Coding Distinction Between DATAPANEL and DATALATTICE

The primary difference between coding the DATAPANEL and DATALATTICE layouts is the way that the classification variables are declared.

DATAPANEL takes one list of variables in parentheses. The number of class variables in the list is unlimited, though the effectiveness of the graph decreases as the number of class variables exceeds three or four. In such a case, it is better to use two class variables, and use the other class variables in the BY statement of the SGRENDER procedure.

layout datapanel classvars=(product division) / . . . ;

DATALATTICE, on the other hand, takes one variable for a row dimension and/or one variable for a column dimension:

layout datalattice rowvar=product colvar=division /  . . .;

The LAYOUT PROTOTYPE Statement

In both the DATAPANEL and the DATALATTICE blocks, the nested PROTOTYPE layout is similar to an OVERLAY layout, with the following major differences:

Multiple plots can be overlaid, but BARCHART is the only computed plot that can be included in the prototype. This means that you cannot use BOXPLOT, DENSITYPLOT, ELLIPSE, HISTOGRAM, REGRESSIONPLOT, LOESSPLOT, PBSPLINE, or MODELBAND statements in the PROTOTYPE layout. See Using Non-computed Plots in Classification Panels for examples of how to work around this limitation.
DISCRETELEGEND, CONTINUOUSLEGEND, ENTRY, ENTRYTITLE, and ENTRYFOOTNOTE statements cannot be included in the PROTOTYPE layout, nor can nested layouts. For information about adding a legend or other information outside of the cells, see Using Sidebars.
Axis options for classification panels are specified on the LAYOUT DATALATTICE or LAYOUT DATAPANEL statement, not on the LAYOUT PROTOTYPE statement. For information about setting axis options for the layout, see Setting Panel Axis Features.