How to plot using PROC SGPLOT from 2 separate PROC MIXED outputs

[jeff_b]

Hello,

I'm trying to plot 2 separate PROC MIXED outputs in 1 PROC SGPLOT.  Here is the output I am keeping -only estimate and 95% CI's. Could you please suggest a way for me to plot them in the same PROC SGPLOT series plot along with 95% CI's?

/*1st output for the total sample*/

proc mixed data = temp;

     class ID visit group;
     model outcome = visit group visit *group / cl ;
     repeated visit / type=un subject=ID r;
     lsmeans visit group visit *group / cl;
    ods output lsmeans = plot1(where=(Effect="visit*group"));
run;

/*2 nd output from a subset of people */

proc mixed data = temp;

    where subset = 1;
     class ID visit group;
     model outcome = visit group visit *group / cl ;
     repeated visit / type=un subject=ID r;
     lsmeans visit group visit *group / cl;
    ods output lsmeans = plot2(where=(Effect="visit*group"));
run;

*This scatter plot plots a 1 output but can add 95% CI's, while proc sgplot series cannot add 95% CI's.

proc sgplot data=plot1;
    scatter y=Estimate x=visit/ yerrorlower=Lower yerrorupper=Upper
    markerattrs=(symbol=diamondfilled) group=GRP groupdisplay=cluster;
    refline 0 / axis=x;
   xaxis grid values=(1 to 3 by 1);
   yaxis values=(7 to 12 by 1) grid colorbands=odd display=(nolabel) discreteorder=data;
run;

[ballardw]

You would combine the data in a data step. Something like this perhaps:

Data allplot;
   set plot1 plot2 indsname=ds;
   Source=ds;
run;

The Indsname option creates a temporary variable with the name of the data set records come from. The Source variable keeps that so you can use it later. You may want to use this as a Group option variable in plots to indicate just which plot you are looking at.

Large economy sized hint: If you use a SERIES plot and a SCATTER plot with the same x and y variables they will overlay each other. So the errorbars from the scatter appear with to be from the series.

[jeff_b]

Hi @ballardw ! Thank you for the quick reply to my post.

I followed your suggestion and combined the 2 data in a data step and plotted them in PROC SGPLOT using both scatter and series. Please see the code below. However, the plot looks a bit messy. I was hoping to get more of connected lines like in time series and 95% CI's as bands around them. Thanks in advance for any suggestions!

proc sgplot data=allplot noautolegend;
    series x=VISIT y=Estimate / group=Source;
    scatter x=VISIT y=Estimate / group=Source yerrorlower=Lower yerrorupper=Upper groupdisplay=cluster;
    xaxis grid values=(1 to 3 by 1);
   yaxis values=(7 to 12 by 1) grid colorbands=odd display=(nolabel) discreteorder=data;
run;

[ballardw]

---

@jeff_b wrote:

Hi @ballardw ! Thank you for the quick reply to my post.

I followed your suggestion and combined the 2 data in a data step and plotted them in PROC SGPLOT using both scatter and series. Please see the code below. However, the plot looks a bit messy. I was hoping to get more of connected lines like in time series and 95% CI's as bands around them. Thanks in advance for any suggestions!

proc sgplot data=allplot noautolegend;
    series x=VISIT y=Estimate / group=Source;
    scatter x=VISIT y=Estimate / group=Source yerrorlower=Lower yerrorupper=Upper groupdisplay=cluster;
    xaxis grid values=(1 to 3 by 1);
   yaxis values=(7 to 12 by 1) grid colorbands=odd display=(nolabel) discreteorder=data;
run;

---

The BAND plot instead of series or scatter then.

proc sgplot data=allplot noautolegend;
    band x=VISIT y=Estimate  lower=Lower upper=Upper
        / group=Source ;
    xaxis grid values=(1 to 3 by 1);
    yaxis values=(7 to 12 by 1) grid colorbands=odd display=(nolabel) discreteorder=data;
run;  

But I think you need to look at your output data set a bit. I'm not sure about that your "limits" are as intended.

[jeff_b · Accepted solution]

Hi @ballardw ! Thanks for the suggestions! A combination PROC SGPLOT with both bands and scatter worked well.

proc sgplot data=allplot;
     band x=visit lower=Lower upper=Upper / group=Source fillattrs=GraphConfidence1 legendlabel="95% CL"       name="band";
     scatter x=visit y=Estimate / group=Source yerrorlower=Lower yerrorupper=Upper groupdisplay=overlay;
      position=bottomright;
      xaxis grid values=(1 to 3 by 1);
      yaxis grid  colorbands=odd display=(nolabel) discreteorder=data;
run;