topic Boxplot in SAS GTL in Graphics Programming

Boxplot in SAS GTL

SASSLICK001 — Mon, 16 Nov 2015 20:30:20 GMT

Hi All

I am trying to produce boxplot using SAS GTL and ended up getting the following message in the log

"WARNING: The data for boxes is incomplete or invalid. Some boxes will not be drawn."

Can someone help me how to get rid of this warning message!

Thanks

John

Re: Boxplot in SAS GTL

Reeza — Mon, 16 Nov 2015 20:34:03 GMT

You'd have to post your code, and preferably some sample data that will generate the error.
What version of SAS are you on?

Re: Boxplot in SAS GTL

SASSLICK001 — Mon, 16 Nov 2015 20:47:20 GMT

I am using SAS 9.4 TSL1M3

Re: Boxplot in SAS GTL

SASSLICK001 — Mon, 16 Nov 2015 20:58:48 GMT

TRTA	TRTAN	AVAL
A	1	5.265563
A	1	2.416991
A	1	6.361984
A	1	2.733437
A	1	4.215147
A	1	4.298788
A	1	3.76999
A	1	3.447105
A	1	3.17307
A	1	4.58969
A	1	5.020884
A	1	5.169635
A	1	2.244204
A	1	2.13761
A	1	2.61073
A	1	2.964504
A	1	3.140009
A	1	5.143698
A	1	4.861378
A	1	3.838215
A	1	2.769648
A	1	3.704952
B	2	3.105561
B	2	2.574498
B	2	6.451404
B	2	3.030039
B	2	2.99401
B	2	4.847858
B	2	3.997044
B	2	2.420613
B	2	5.188032
B	2	3.510903
B	2	3.144169
B	2	3.79773
B	2	3.302072
B	2	2.952558

proc template;
define statgraph box;
begingraph/ designwidth=9.3in designheight=4.5in border=yes;;
entrytitle "X" / textattrs=(size=9);

layout lattice / columns=1 rows=1 columnweights=(0.7) ;

layout overlay / yaxisopts=(linearopts=(tickvaluesequence=(start=1 end=7 increment=1) viewmin=1 viewmax=7) label='X')
xaxisopts=(display=(tickvalues label) label="Treatment Group");

boxplot x=trta y=eval(ifn(trtan=1,aval, .)) / whiskerpercentile=10 display=(mean median outliers caps fill) capshape=line
discreteoffset=0 boxwidth=.27 name='a' legendlabel="A"
outlineattrs=(color=blue thickness=2) outlierattrs=(color=blue symbol=square /*trianglefilled*/)
meanattrs=(color=blue symbol=asterisk)
medianattrs=(color=blue thickness=2)
whiskerattrs=(color=blue thickness=2);

boxplot x=trta y=eval(ifn(trtan=2,aval, .)) / whiskerpercentile=10 display=(mean median outliers caps fill) capshape=line
discreteoffset=0 boxwidth=.27 name='b' legendlabel="B"
outlineattrs=(color=red thickness=2) outlierattrs=(color=red symbol=square)
meanattrs=(color=red symbol=asterisk)
medianattrs=(color=red thickness=2)
whiskerattrs=(color=red thickness=2);
endlayout;
endlayout;
endgraph;
end;
run;

proc sgrender data=adpp template=box;;
run;

Re: Boxplot in SAS GTL

DanH_sas — Mon, 16 Nov 2015 21:25:25 GMT

You're attempting to use the ifn() expression to "where out" different category values so you treat the data as overlays instead of group data. This makes boxes with all missing values, causing the messages to be generated. It appears you're doing this to control the attributes for each box. A cleaner way to do this (which will also eliminate your messages) is to remove the ifn()'s and use attributes maps instead. That way, you can control the attributes for each group. If you need more information on attribute maps, look in this paper I wrote back in 2014: http://support.sas.com/resources/papers/proceedings14/SAS156-2014.pdf.

Hope this helps!

Dan

Re: Boxplot in SAS GTL

DanH_sas — Tue, 17 Nov 2015 13:47:52 GMT

Using the data from your example, the code for the solution I proposed is below. I used an attrmap data set instead of putting the attrmap directly in the template, as it gives you more flexibility and re-use for other templates you might use for rendering the same treatment groups.

data attrmap;
retain id "trtgrp";
length linecolor $ 4;
input value $ linecolor $;
cards;
A blue
B red
;
run;

proc template;
define statgraph box;
begingraph/ designwidth=9.3in designheight=4.5in border=yes;;
entrytitle "X" / textattrs=(size=9);
layout lattice / columns=1 rows=1 columnweights=(0.7) ;
layout overlay / yaxisopts=(linearopts=(tickvaluesequence=(start=1 end=7 increment=1) viewmin=1 viewmax=7) label='X')
xaxisopts=(display=(tickvalues label) label="Treatment Group");
boxplot x=trta y=aval / whiskerpercentile=10 display=(mean median outliers caps fill) capshape=line
name='a' outlineattrs=(thickness=2) outlierattrs=(symbol=square /*trianglefilled*/) group=TRTA

groupdisplay=cluster meanattrs=(symbol=asterisk) medianattrs=(thickness=2) whiskerattrs=(thickness=2);
endlayout;
endlayout;
endgraph;
end;
run;
ods html;

proc sgrender data=adpp template=box dattrmap=attrmap;
dattrvar TRTA="trtgrp";
run;