box not plotting with 2 y axes

[Manj]

yes , this works. but wonder why my code which is on the similar lines doesnt work

[quickbluefish]

Interesting @DanH_sas - it looks like you *did* apply jitter to the ageatstart variable here?

[DanH_sas]

No, I did not apply any data jittering here. I think you hit the correct root cause. 

[quickbluefish · Accepted solution]

Only other thing I can think of is that the 'jitter' you're applying is causing some sort of issue. If you remove the jitter, do you get the same problem? When I test this on fake data, I get warnings in the log saying that not all boxes were drawn if I either apply jitter OR if I specify 2 separate y-axes. I suspect SGPLOT might not be capable of this particular combination of things.

[Quentin]

If you comment out the xaxis statement, do you get boxes?

If you comment out yaxis statement?

I wonder if the values you entered on the xaxis statement are not the actual values in your dataset.   Or if your values are outside the yaxis.

One thing I don't like about SGPLOT vs the much older traditional GPLOT is that SGPLOT does not throw a NOTE when you have data outside of your axes.  It's possible all your data is outside the axes.

You didn't show your MYATTR dataset.  Could it be everything is set to transparent?

I would suggest start with a working plot that is simple, maybe:

proc sgplot data=ip ;
  vbox aval1 / category=vis group=param1 nomean nooutliers whiskerpct=25 
  connect=median lineattrs=( pattern=1) attrid=param1;
run;

And then add back more of your real code incrementally, until the boxes disappear.

[Manj]

Looks like the jitter created the issue. Thank you so much for pointing this.

[Quentin]

In the DATA step you manually introduced jitter in the values of VIS.

In the XAXIS statement you specified the values to be plotted:

values=(0,1,2,3,4,6,8,10,12,16,20,24,28,32,36,40,44,48,52 )

My guess is after you jitter the data, those values don't appear in your data.  

[Manj]

okay !, but I had done it for line plots like this earlier. May be works different for box plots

[Quentin]

I think it all depends on your axis. For series plots the x-axis is typically continuous, so you can do a bit of jittering. But for box plots, the x-axis is categorical. So if you do the jittering and specify values for the x-axis, you would need to specify jittered values on the x axis.

[Manj]

Thank you both, I made the plot as below, however i face an issue with inverted whisker line on some of the boxes. why is this appearing and how to remove this? Thanks

proc sgplot data=ip dattrmap=myattr;

vbox aval1 / category=avisit group=param lineattrs=( pattern=1) nomean whiskerpct=25 nooutliers connect=median nofill whiskerattrs=(thickness=0.00001) attrid=param1 name='v1';
vbox aval2 / category=avisit group=param y2axis nomean whiskerpct=25 nooutliers connect=median nofill attrid=param1 name='v2';

yaxis label='Normalized AAA value' values=(0 to 1.3 by 0.1) valueattrs=(size=8);
y2axis label='The reciprocal antibody titer' TYPE=LOG LOGBASE=2 valueattrs=(size=8) values=(6.25,12.5,25,50,100,200,400,800,1600,3200,6400,12800,25600,51200,102400,204800,1638400,3276800 ) valuesformat=best. ;
xaxis label='Analysis visit' DISCRETEORDER=DATA ;*values=(0,1,2,3,4,6,8,10,12,16,20,24,28,32,36,40,44,48,52 ) valuesdisplay=('Baseline' 'Week 1' 'Week 2' 'Week 3' 'Week 4' 'Week 6' 'Week 8' 'Week 10' 'Week 12' 'Week 16' 'Week 20' 'Week 24' 'Week 28' 'Week 32' 'Week 36' 'Week 40' 'Week 44' 'Week 48' 'Week 52' ) fitpolicy=rotate valuesrotate=diagonal valueattrs=(size=8);

keylegend "v1" / title="" position=bottom location =outside down=1 valueattrs=(size=8) noborder;

run;

[Quentin]

That's odd.  If your goal is to get rid of the whiskers, you could try:

whiskerattrs=(thickness=0)

Suggestion taken from:

https://communities.sas.com/t5/Graphics-Programming/Creating-a-boxplot-without-whiskers/td-p/248745

[DanH_sas]

What is your goal in setting WHISKERPCT=25? Are you trying to make the whiskers go away?