Using PROC SGPANEL to Create Classification Panels

When creating a panel like the one shown in Using Non-computed Plots in Classification Panels, you might find it easier to create the panel by using PROC SGPANEL in SAS/GRAPH because the procedure does all the necessary data computations for you. For example, the REGRESSIONPLOT, LOESSPLOT, and PBSPLINEPLOT statements have been incorporated into the SGPANEL procedure as REG, LOESS, and PBSPLINE statements. (SGPANEL can also generate other plot types.) By default on PROC SGPANEL, the PANELBY statement creates a DATAPANEL layout.

ods listing style=statistical;

title "Predicted Response to Dosage (mg) over Time";
proc sgpanel data=trial;
  panelby dose / rows=1;
  reg x=days y=response / cli clm;
run;

Classification Panel Created with PROC SGPANEL

Most, but not all, features of the DATALATTICE and DATAPANEL layouts are provided in the SGPANEL procedure.

The SGPANEL procedure supports computed plot statements such as HISTOGRAM, DENSITY, DOT, VBOX, and HBOX (vertical and horizontal box plots). The PANELBY statement controls the layout, determining whether a DATAPANEL, DATALATTICE, or other layout is used to produce the graph. ROWAXIS and COLAXIS statements control the external axes, and the KEYLEGEND statement creates legends, which are placed in sidebars for you.

The SGPANEL procedure does not have a PROTOTYPE block because all of the plot statements after PANELBY are considered part of the prototype. The SGPANEL procedure generates GTL template code behind the scenes and executes the template to create its output. See the SAS/GRAPH Statistical Graphics Procedures Guide documentation for details.

The following example shows additional features of SGPANEL:

ods listing style=statistical;

title "Cholesterol Distribution by Gender and Weight";
proc sgpanel data=sashelp.heart;
  panelby sex weight_status /  layout=lattice onepanel novarname;
  histogram cholesterol;
  density cholesterol / name="density";
  refline 227 / axis=x name="ref" legendlabel="Overall Mean = 227";
  rowaxis  offsetmin=0 offsetmax=.1 max=30;
  keylegend  "density" "ref";
run;