SGPANEL strange behaviour

[USchu]

Hi everybody.

I faced a strange result when producing some slighty complicated sgpanel plots and tried to simplify the example as far as possible.

Running the following program, SAS results in a 200% bar for the  graph. 

DATA test;  

 /*<1>*/  varX=.; varY=2; varZ=7; OUTPUT;

 /*<2>*/  varX=1; varY=1; varZ=1; OUTPUT;

 /*<3>*/  varX=1; varY=1; varZ=2; OUTPUT;

RUN;

PROC SGPANEL DATA= test PCTLEVEL=GROUP; 

PANELBY varX;

  VBAR varY / GROUP=varZ STAT=PERCENT;

ROWAXIS LABEL='% of Z';  ;

RUN;

However, removing the Missing VarX dataline (<1>) or giving allways varZ=1 vor varX=1  (dataline<3>), sgpanel works fine.

Does anyone have an explanation for this?

I just can solve it by removing any varX=. lines using

Proc SGPANEL DATA=test(WHERE= (varX^=.));

but since this is simplified from a rather complicated multiuse macro, I would prefere to know the background of this result.

Thanks for any help.

Uli

[DanH_sas]

This does appear to be an issue that needs to be investigated. I would recommend entering a ticket to Technical Support with this simple example you've created. In the meantime, you have two choices:

1. If you want to exclude missing panel variable values from your display, WHERE clause them out as you propose.
2. If you are okay with seeing these panels with missing values, just add the MISSING option to the PANELBY statement, and you should get the correct result.

Hope this helps!

[USchu]

Thank you, Dan,

for your valuable answer. I think I will follow your advice and write a ticket to Technical Support.

Best regards Uli

[Ksharp]

Is what you are looking for ?

DATA test;  
 /*<1>*/  varX=.; varY=2; varZ=7; OUTPUT;
 /*<2>*/  varX=1; varY=1; varZ=1; OUTPUT;
 /*<3>*/  varX=1; varY=1; varZ=2; OUTPUT;
RUN;
 
PROC SGPANEL DATA= test PCTLEVEL=CELL; 
PANELBY varX;
  VBAR varY / GROUP=varZ STAT=PERCENT ;
ROWAXIS LABEL='% of Z';  ;
RUN;

[USchu]

Thank you, Ksharp,

for your valuable proposal. I am sorry, my sample might have been to much simplified. I need indeed group percentages. However, the problem can be solved by the WHERE-clause mentioned above. I just wanted to know the background of this behaviour in order to be save using the program.

best regards Uli