GTL lattice nested in datapanel

[jessek]

I have multiple elements in a lattice layout, but I want them repeated based on a classification variable.  Is there any way to do this?  I tried just putting them inside a datapanel layout or datalattice layout, but it doesn't work.

For example:

This code produces a bar char with information on either side.  These values are all from one year.  What if I had 4 years of values and I wanted to produce the same output 4 times, once for each year based all in the same graphic.  Is there a way to do this?

data PGA2007;

  input Rank 2.  Player & $15. Age Events Rounds CutsMade Top10 Wins Earnings;

  retain Constant 1;

  label CutsMade="Cuts Made" Top10="Top 10s";

  format age events wins 3.0;

  datalines;

1 Tiger Woods     33  16  61 16 12  7 10867052

2 Phil Mickelson  38  22  73 16  7  3  5819988

3 Vijay Singh     45  27 101 25  7  2  4728377

4 Steve Stricker  41  23  80 19  9  1  4663077

5 K.J. Choi       38  25  88 20  7  2  4587859

6 Rory Sabbatini  32  23  80 18 10  1  4550040

7 Jim Furyk       38  24  84 20  8  1  4154046

8 Zach Johnson    32  23  78 18  5  2  3922338

9 Sergio Garcia   29  19  67 16  7  0  3721185

10 Aaron Baddeley  27  23  82 19  7  1  3441119

;

run;

proc template;
  define statgraph BarTableHorz;
  begingraph;
  entrytitle "Professional Golf Statistics for 2007";  
  layout lattice / columns=4 columngutter=2
                   columnweights=(.06 .06 .82 .06);
    column2headers;
      entry halign=right "Age"    / textattrs=(size=7pt weight=bold);
      entry halign=right "Events" / textattrs=(size=7pt weight=bold);
      entry halign=right ""   / textattrs=(size=7pt weight=bold); 
      entry halign=right "Wins"   / textattrs=(size=7pt weight=bold); 
    endcolumn2headers;

    /* columns to left of barchart */
    layout overlay / walldisplay=none border=false
                     yaxisopts=(reverse=true type=discrete display=none)
                     xaxisopts=(display=none offsetmin=0.3 offsetmax=0);
      scatterplot y=player x=constant  / markercharacter=age markerattrs=(size=0) ;
    endlayout;

    layout overlay / walldisplay=none border=false
                     yaxisopts=(reverse=true type=discrete display=none)
                     xaxisopts=(display=none offsetmin=.3 offsetmax=0);
      scatterplot y=player x=constant  / markercharacter=events markerattrs=(size=0);
    endlayout;

    /* barchart with inset */

    layout overlay / yaxisopts=(reverse=true display=(tickvalues) tickvalueattrs=(weight=bold))
                     xaxisopts=(griddisplay=on labelattrs=(weight=bold)
                                linearopts=(tickvalueformat=(extractscale=true)) );
       
      barchart x=player y=Earnings / barlabel=true barlabelformat=dollar12. orient=horizontal
                                       outlineattrs=(color=black);

endlayout;
     layout overlay / walldisplay=none border=false
                     yaxisopts=(reverse=true type=discrete display=none)
                     xaxisopts=(display=none offsetmin=.3 offsetmax=0);
        scatterplot y=player x=constant  / markercharacter=wins  markerattrs=(size=0);
   

     endlayout;
  endlayout;

  endgraph;
  end;
run;

ods listing close;
ods html image_dpi=100 file='BarTableHorz.html' path='.';
ods graphics / reset noborder width=700px height=400px
  imagename='BarTableHorz' imagefmt=gif noscale;

proc sgrender data=PGA2007 template=BarTableHorz;
run;

ods html close;
ods listing;

Thanks for any help.

-Jesse

[Jay54 · Accepted solution]

Each cell in a Layout DataLattice can only be a single cell graph defined in the LAYOUT PROTOTYPE statement.  Layout Prototype is essentially a Layout Overlay.  You cannot use a multi-cell graph as a prototype in a data lattice.  Of course, you can run your Lattice graph above with a BY variable, and get one graph for each year.  The main benefit of a Datalattice are the uniform axes across years.

However, if you want to create a prototype that is essentially one graph, and some data columns like the Golf example, that can be done by using the "axis-splitting" technique that I have described in this paper:

http://support.sas.com/resources/papers/proceedings11/281-2011.pdf.  See #4 Forest Plot.

Here is an example.  In this example, I have added a variable YEAR to the data, that has two values 2007 and 2008.  You can have more values.  Then, I have defined a single cell prototype with one bar chart and multiple scatterplots to do the statistics table using the X and X2 axis-splitting technique.  Then, I used a layout DataLattice with ROWVAR=Year.   With this technique, all the stat columns have to go on one side.  If this is what you are looking for, I can clean up and attach the program.

Note, you have to set OFFSETMIN and OFFSETMAX on the COLUMNAXISOPTS and COLUMN2AXISOPTS appropriately based on how many statistics columns you have.

Program:

data PGA; 
  input Year Rank  Player $ 9-24  Age Events Rounds CutsMade Top10 Wins Earnings;
  retain Constant 1;
  label CutsMade="Cuts Made" Top10="Top 10s";
  format age events wins 3.0;
  Agelabel='Age';
  EventsLabel='Events';
  WinsLabel='Wins';
  EarningsLabel='Earnings';
  datalines;
2007  1 Tiger Woods      33  16  61 16 12  7 10867052
2007  2 Phil Mickelson   38  22  73 16  7  3  5819988
2007  3 Vijay Singh      45  27 101 25  7  2  4728377
2007  4 Steve Stricker   41  23  80 19  9  1  4663070
2007  5 K.J. Choi        38  25  88 20  7  2  4587859
2007  6 Rory Sabbatini   32  23  80 18 10  1  4550040
2007  7 Jim Furyk        38  24  84 20  8  1  4154046
2007  8 Zach Johnson     32  23  78 18  5  2  3922338
2007  9 Sergio Garcia    29  19  67 16  7  0  3721185
2007 10 Aaron Baddeley   27  23  82 19  7  1  3441119
;
run;

proc sort data=pga out=pga2;
  by descending rank;
  run;

 
/*--Simulate and add data for Year=2008--*/
data pga3;
  set pga2;
  output;
  Year=2008;
  Rank=mod(rank+2, 10)+1;
  Age=age+1;
  Events=events*(1+0.2*(ranuni(2)-0.5));
  Rounds=rounds*(1+0.2*(ranuni(2)-0.5));
  CutsMade=cutsmade*(1+0.2*(ranuni(2)-0.5));
  Top10=top10*(1+0.2*(ranuni(2)-0.5));
  Wins=wins*(1+0.5*(ranuni(2)-0.5));
  Earnings=earnings*(1+0.2*(ranuni(2)-0.5));
  output;
 run;

/*--Create template for graph--*/
proc template;
  define statgraph BarTableHorz;
  begingraph;
    entrytitle "Professional Golf Statistics";   
    layout datalattice rowvar=year / columnaxisopts=(offsetmin=0 offsetmax=0.38 griddisplay=on
                                     linearopts=(tickvalueformat=(extractscale=true)))
                    column2axisopts=(offsetmin=0.7 offsetmax=0.05 display=(tickvalues));
      layout prototype;
        barchart x=player y=Earnings / orient=horizontal skin=modern;
        scatterplot y=player x=earningslabel / markercharacter=earnings xaxis=x2;
        scatterplot y=player x=agelabel / markercharacter=age xaxis=x2;
        scatterplot y=player x=eventslabel / markercharacter=events xaxis=x2;
        scatterplot y=player x=winslabel / markercharacter=wins xaxis=x2;
      endlayout;
    endlayout;
  endgraph;
  end;
run;

/*--Render the graph--*/
ods graphics / reset noborder width=8in height=5in imagename='GolfLattice';
proc sgrender data=PGA3 template=BarTableHorz;
format earnings dollar12.0;
run;

[jessek]

Thanks so much for figuring out a solution for me.  This will work perfectly.  Hopefully I will be able to apply the information in your paper to my project, but it would be great if you could also send/post the code used in the simple golf example.  Thanks again!

-Jesse

[Jay54]

Program added.