topic How to sort xaxis values by yaxis values in sgplot? in SAS Procedures

How to sort xaxis values by yaxis values in sgplot?

emaneman — Wed, 16 Feb 2022 10:59:40 GMT

Dear all,

I am using sgplot to plot studentresid using the following code


proc sgplot data=second ;
scatter y=studentresid x=id / jitter transparency=0.5 
markerattrs=(symbol=CircleFilled size=12) ;

yaxis values=(3 to 3 by 1) grid;

And I obtain the attached scatter plot. However, I would like to improve it by:

- sorting the xaxis by ascending values of the studentresid, from left to right

- ideally coloring differently the three observations with very high studentresid, above the grid line at value 3

Re: How to sort xaxis values by yaxis values in sgplot?

Rick_SAS — Wed, 16 Feb 2022 11:17:10 GMT

Sort by the rank of the Y variable and add an indicator variable that colors the markers for which studentized resid is larger than 3.

/* create fake data */
data second;
call streaminit(1);
do id = 1 to 220;
   studentresid = rand("Normal", 2, 0.1);
   if id=4 then studentresid=6;
   else if id=55 then studentresid=5;
   else if id=130 then studentresid=4;
   output;
end;
run;

/* sort according to rank */
proc rank data=second out=Want1 descending;
var studentresid;
ranks Rank;
run;

proc sort data=Want1; by Rank; run;

/* add indicator variable for outlier */
data Want;
set Want1;
Outlier = (studentresid > 3);
run;

proc sgplot data=Want;
scatter y=studentresid x=Rank / jitter transparency=0.5 GROUP=Outlier
markerattrs=(symbol=CircleFilled size=12) ;
yaxis values=(3 to 3 by 1) grid label="Studentized Residuals";
xaxis display=(nolabel);
run;

Re: How to sort xaxis values by yaxis values in sgplot?

Rick_SAS — Wed, 16 Feb 2022 11:35:18 GMT

I just re-read the OP and noticed that the request was for ASCENDING values of studentized residuals. To get that, remove the DESCENDING keyword from the PROC RANK statement.

Re: How to sort xaxis values by yaxis values in sgplot?

emaneman — Wed, 16 Feb 2022 12:07:21 GMT

Thank you Rick! Fast, brilliant, and precise, as usual!