SGPLOT with VBAR percent labels and YAXIS with response sum labels

[subpar_actuary]

Hi All,

Apologies if this has been answered. I've tried to find it in the topics to no avail... maybe not using the right key words?

Thanks to @Reeza for helping me realize I should switch to SGPLOT from using GCHART pretty much exclusively until now.

My problem is that I want to be able to do something in SGPLOT that I can do in GCHART, which is show percent labels above the VBARs, but show the SUMVAR sum on the Y Axis. So far in SGPLOT I have been able to get percent over both VBAR and on YAXIS, or response sums over VBAR and on YAXIS.

Below is some code that shows first what I want to do in SGPLOT (but using GCHART as example), then second and third where I am so far with SGPLOT:

proc format;
	value HPBins
		low -< 100  = '<100'
		100 -< 200  = '<200'
		200 -< 300  = '<300'
		300 -< 400  = '<400'
		400 -  high = '400+';
run;

/* FIRST */
title1 'What I want to show in SGPLOT, using GCHART as example';
title2 'GCHART: Percent label over VBAR, sumvar SUM on Y-Axis';
proc gchart data=SASHELP.CARS;
	format Horsepower HPBins.;
	vbar Horsepower / discrete sumvar=MSRP outside=percent;
run;

/* SECOND */
title1 'Not quite there...';
title2 'SGPLOT: Response SUM labels over VBAR and on Y-Axis';
proc sgplot data=SASHELP.CARS;
	format Horsepower HPBins.;
	vbar Horsepower / response=MSRP datalabel;
	xaxis values=('<100' '<200' '<300' '<400' '400+');
	run;
run;

/* THIRD */
title1 'Not quite there...';
title2 'SGPLOT: Percent labels over VBAR and on Y-Axis';
proc sgplot data=SASHELP.CARS;
	format Horsepower HPBins.;
	vbar Horsepower / response=MSRP stat=percent datalabel;
	xaxis values=('<100' '<200' '<300' '<400' '400+');
	run;

[Reeza · Accepted solution]

Someone also asked this recently and I posted a solution....if you search through my post history you'll find it. The trick is to do it twice and suppress one plot or pre-summarize and customize the label with a different variable.

[ballardw]

One way that gets pretty close

proc sgplot data=SASHELP.CARS noautolegend;
	format Horsepower HPBins.;
   vbar horsepower / response=msrp stat=sum name='sum'

                     ;
	vbar Horsepower / response=MSRP stat=percent 
                    datalabel y2axis name='perc' ;
	xaxis values=('<100' '<200' '<300' '<400' '400+');
run;

Personally I never had good luck with Gchart and created my own summary values so I could explicitly state variables and/or values.

Which with SGPLOT and overlay of Textplot or similar provides a lot of control.

[subpar_actuary]

@Reeza thanks I found the post.

Pretty nifty way to rig the graph. I had some problems with the back (first) vbar showing up in the background, so I added transparency=1 which fixed that. I assume that issue is because the left and right axes are different bases, and since SAS fits each axis to nice round numbers, they sometimes don't line up horizontally.  

Running @ballardw 's code and adding 'by make', you can see the issue with the Dodge graph and a number of others:

Here's the final code that adds the transparency option (along with other options to get percents at by level, remove the y2axis labels, and change the ugly orange to blue 😝. Thanks all for your help!

proc sgplot data=SASHELP.CARS noautolegend pctlevel=by;
	format horsepower HPBins.;
	vbar horsepower / response=msrp stat=sum
		name='sum' transparency=1;
	vbar Horsepower / response=msrp stat=percent
		datalabel y2axis name='perc' fillattrs=(color=vligb) outlineattrs=(color=bib);
	xaxis values=('<100' '<200' '<300' '<400' '400+');
	y2axis display=none;
	by make;
	run;

[ballardw]

You could use VALUES=() lists on both axis to restrict the range of values, make things align and should address the "overlap"/"peek-a-bou" issue arising from the SAS underlying rules for building the axis range, or possibly as simple as MAX= on each axis.