PROC SGPANEL

[Recep]

Hello there,

I'm trying to create a graph with SGPLOT that shows rates by hospitals in two seperate panels: Level_1_2 and Level_3. With the dataset I'm currently using, the problem is, I cannot divide the panels the way I want it. The final graph is supposed to have the rates for each hospital within their "level". Currently the graph I produce includes Level_3 hospitals in the Level_1_2 panel and Level_1_2 hospitals in the Level_3 panel. Do you think I set up my dataset not "optimally" or am I missing something in the SGPANEL?

Also, if its a possibility I would like to create two seperate REFLINES for each panel.

Thanks a lot for your help in advance!

Here are my dataset, code, and the graph.

hosp_name	level	cpred_final	clower_final	cupper_final
VGH	Level_1_2	27.8587	26.0588	29.7581
VicGH	Level_1_2	21.7479	19.824	23.8298
RCH	Level_1_2	19.2273	17.5596	21.0323
KGH	Level_1_2	19.0601	16.8262	21.5502
RIH	Level_1_2	17.7454	15.5962	20.1532
UHNBC	Level_3	13.5041	11.1077	16.3623
NRGH	Level_3	10.9852	9.0805	13.2551
LGH	Level_3	10.9797	9.2928	12.9473
ARH	Level_3	10.4652	8.5618	12.757
SPH	Level_3	8.8174	6.8647	11.2857

proc sgpanel data=Results noautolegend;
panelby level/columns=2 novarname;
rowaxis values=(0 to 30 by 5) grid;
colaxis grid;
scatter x=hosp_name y=cpred_final / markerattrs=(symbol=circle color=red size=8)
DATALABEL=cpred_final
DATALABELATTRS=(color=black /*weight=bold*/)
DATALABELPOS=right
yerrorlower=clower_final
yerrorupper=cupper_final;

refline 21.13 10.95/label=('Level 1&2 average: 21.1%' 'Level 3 average: 11.0%');
title h=12pt justify=left bold italic 'Figure 2: Adjusted predictions at the average risk score by hospital';
format cpred_final 4.1;
run;

[DanH_sas · Accepted solution]

On the PANELBY statement, add UNISCALE=ROW after "/" and you should get what you want. Basically, you only want the ROW axis to be uniform, but both ROW and COLUMN axes are uniform by default, giving you the combined tick list.

Hope this helps!

Dan

[Recep]

This is great Dan! Yes, it resolved my problem. Do you think I can create two seperate reflines for each panel?

Cheers,

Recep

[DanH_sas]

Absolutely! Instead of putting your refline values on the REFLINE statement (which will replicate in every cell), but the values in a column and reference the colume from the REFLINE statement:

hosp_name	level	cpred_final	clower_final	cupper_final	ref_level	ref_label
VGH	Level_1_2	27.8587	26.0588	29.7581	21.13	Level 1&2 average: 21.1%
VicGH	Level_1_2	21.7479	19.824	23.8298	.
RCH	Level_1_2	19.2273	17.5596	21.0323	.
KGH	Level_1_2	19.0601	16.8262	21.5502	.
RIH	Level_1_2	17.7454	15.5962	20.1532	.
UHNBC	Level_3	13.5041	11.1077	16.3623	10.95	Level 3 average: 11.0%
NRGH	Level_3	10.9852	9.0805	13.2551	.
LGH	Level_3	10.9797	9.2928	12.9473	.
ARH	Level_3	10.4652	8.5618	12.757	.
SPH	Level_3	8.8174	6.8647	11.2857	.

proc sgpanel data=Results noautolegend;
panelby level/columns=2 novarname uniscale=row;
rowaxis values=(0 to 30 by 5) grid;
colaxis grid;
scatter x=hosp_name y=cpred_final / markerattrs=(symbol=circle color=red size=8)
DATALABEL=cpred_final
DATALABELATTRS=(color=black /*weight=bold*/)
DATALABELPOS=right
yerrorlower=clower_final
yerrorupper=cupper_final;
refline ref_level / label=ref_label
title h=12pt justify=left bold italic 'Figure 2: Adjusted predictions at the average risk score by hospital';
format cpred_final 4.1;
run;

[Recep]

Thanks a lot Dan! This also resolve it...

Have a good day...

[ballardw]

The hospital labels are the same because the XAXIS for sgpanel by default is the same for each panel.

You might consider taking that output and using it as the basis for the ODS Graphics Designer to build the custom template that would be needed for SGRender to build what I think you are meaning to do.