Simple way to label your boxplot?

[Quentin]

I think you need to keep the variable SEX in work.temp, and add group=sex to your scatter plots.  I also added groupdisplay=cluster and clusterwidth.  This is not working, but I would play with code like this:

data have;
 set sashelp.heart(obs=1000);
run;

ods select none;
ods output sgplot=sgplot1;
proc sgplot data=have  ;
vbox weight/group=sex category=bp_status ;
run;
ods select all;

%let dsid=%sysfunc(open(sgplot1));
%let first=%sysfunc(nliteral(%sysfunc(varname(&dsid,1))));
%let second=%sysfunc(nliteral(%sysfunc(varname(&dsid,2))));
%let third=%sysfunc(nliteral(%sysfunc(varname(&dsid,3))));
%let dsid=%sysfunc(close(&dsid));



data temp(rename=(&third.=bp_status &first.=_y &second.=_stat));
 set sgplot1;
 _value=round(&first.,0.01);
 if not missing(&second.) and &second. in ('MIN' 'MAX' 'MEDIAN' 'MEAN' 'Q1' 'Q3');
 keep _value &first. &second. &third. sex;
run;
data want;
 set have temp;
 if _stat  in ( 'MEDIAN') then _value1=_value;
 if _stat  in ( 'MIN' 'MAX') then _value2=_value;
 if _stat  in ( 'MEAN' ) then _value3=_value;
run;
title  c=black  "foo";  
proc sgplot data=want noautolegend ;
vbox weight/group=sex category=bp_status nomean boxwidth=0.1  groupdisplay=cluster clusterwidth=.8;
scatter x=bp_status y=_y /markerchar=_value1 labelstrip  markercharattrs=(size=8) group=sex groupdisplay=cluster clusterwidth=.8;
scatter x=bp_status y=_y /markerchar=_value2 labelstrip  markercharattrs=(size=8) group=sex groupdisplay=cluster clusterwidth=.8;
scatter x=bp_status y=_y /markerchar=_value3 labelstrip  markercharattrs=(size=8) group=sex groupdisplay=cluster clusterwidth=.8;

scatter x=bp_status y=_value3 /markerattrs=(color=darkred symbol=circlefilled) name='mean' legendlabel='Mean' group=sex  groupdisplay=cluster clusterwidth=.8;
scatter x=bp_status y=_value1 /markerattrs=(color=navy symbol=diamondfilled) name='median' legendlabel='Median' group=sex  groupdisplay=cluster clusterwidth=.8;

yaxis integer;
keylegend 'mean' 'median'/AUTOITEMSIZE;
run;
title;

[JFuchsia]

Thank you, this was certainly worth a shot! 🙂 

[JFuchsia]

So, 

I've decided to go with SGPANEL and have given up on SGPLOT and got the following code going:

data have;
 set sashelp.heart(obs=1000);
run;

%macro box_plot(dsn=,x=,y=,title=);
ods select none;
ods output sgpanel=sgplot2;
proc sgpanel data=&dsn.;
panelby bp_status;
vbox &y./  category=&x. ;
run;
ods select all;

%let dsid=%sysfunc(open(sgplot2));
%let first=%sysfunc(nliteral(%sysfunc(varname(&dsid,1))));
%let second=%sysfunc(nliteral(%sysfunc(varname(&dsid,2))));
%let third=%sysfunc(nliteral(%sysfunc(varname(&dsid,3))));
%let dsid=%sysfunc(close(&dsid));

data temp(rename=(&third.=&x. &first.=_y &second.=_stat));
 set sgplot2;
 _value=round(&first.,0.01);
 if not missing(&second.) and &second. in ('MIN' 'MAX' 'MEDIAN' 'MEAN' 'Q1' 'Q3');
 keep _value &first. &second. &third. bp_status;
run;
data want;
 set &dsn. temp;
 if _stat  in ( 'MEDIAN') then _value1=_value;
 if _stat  in ( 'MIN') then _value2=_value;
 if _stat  in ('MAX') then _value3=_value;
 if _stat  in ( 'MEAN' ) then _value4=_value;
run;
title  c=black  "&title.";  
proc sgpanel data=want noautolegend ;
panelby bp_status;
vbox &y./category=&x. nomean boxwidth=0.2;

scatter x=&x. y=_y / discreteoffset=0.2  markerchar=_value1 labelstrip  markercharattrs=(size=8);
scatter x=&x. y=_y /discreteoffset=0.2  markerchar=_value2 labelstrip  markercharattrs=(size=8);
scatter x=&x. y=_y /discreteoffset=0.2  markerchar=_value3 labelstrip  markercharattrs=(size=8);
scatter x=&x. y=_y /discreteoffset=0.2  markerchar=_value4 labelstrip  markercharattrs=(size=8);

scatter x=&x. y=_value4 /markerattrs=(color=darkred symbol=circlefilled) name='mean' legendlabel='Mean' ;
scatter x=&x. y=_value1 /markerattrs=(color=navy symbol=diamondfilled) name='median' legendlabel='Median' ;

keylegend 'mean' 'median'/AUTOITEMSIZE;
run;
title;
%mend;

%box_plot(dsn=have, x=sex, y=weight, title=VCSS-Score)

I greatly dislike the way the panel's output in this one as it's hard to tell what the scale is all the time:

I would like everything in one panel but what can I do??

Thank you,

J Fuchsia

[JFuchsia]

I added 

panelby bp_status/layout=COLUMNLATTICE;

And this seemed to fix my concerns!

[Ksharp]

You want this ?

data have;
 set sashelp.heart(obs=1000);
run;
proc sort data=have;
by bp_status sex;
run;


%macro box_plot(dsn=,group=,x=,y=,title=);
ods select none;
ods output sgplot=sgplot1;
proc sgplot data=&dsn.  ;
vbox &y./group=&group. category=&x. ;
run;
ods select all;

%let dsid=%sysfunc(open(sgplot1));
%let first=%sysfunc(nliteral(%sysfunc(varname(&dsid,1))));
%let second=%sysfunc(nliteral(%sysfunc(varname(&dsid,2))));
%let third=%sysfunc(nliteral(%sysfunc(varname(&dsid,3))));
%let forth=%sysfunc(nliteral(%sysfunc(varname(&dsid,4))));
%let dsid=%sysfunc(close(&dsid));

data temp(rename=(&third.=_x &first.=_y &second.=_stat &forth.=_group));
 set sgplot1;
 _value=round(&first.,0.01);
 if not missing(&second.) and &second. in ('MIN' 'MAX' 'MEDIAN' 'MEAN' 'Q1' 'Q3');
 keep _value &first. &second. &third. &forth.;
run;
data want;
 set &dsn. temp;
 if _stat  in ( 'MEDIAN') then _value1=_value;
 if _stat  in ( 'MIN' 'MAX') then _value2=_value;
 if _stat  in ( 'MEAN' ) then _value3=_value;
run;
title  c=black  "&title.";  
proc sgplot data=want noautolegend ;
vbox &y./group=&group. category=&x. nomean boxwidth=0.3 name='box';
scatter x=_x y=_y /group=_group groupdisplay=cluster clusterwidth=0.7 discreteoffset=0.12  markerchar=_value1 labelstrip  markercharattrs=(color=black size=8) ;
scatter x=_x y=_y /group=_group groupdisplay=cluster clusterwidth=0.7 discreteoffset=0.12  markerchar=_value2 labelstrip  markercharattrs=(color=black size=8) ;
scatter x=_x y=_y /group=_group groupdisplay=cluster clusterwidth=0.7 discreteoffset=0.12  markerchar=_value3 labelstrip  markercharattrs=(color=black size=8)  ;

scatter x=_x y=_value3 /group=_group groupdisplay=cluster clusterwidth=0.7 markerattrs=(color=darkred symbol=circlefilled) ;
scatter x=_x y=_value1 /group=_group groupdisplay=cluster clusterwidth=0.7 markerattrs=(color=navy symbol=diamondfilled)  ;

yaxis integer;
legenditem name='mean' type=MARKER / label='Mean' MARKERATTRS=(color=darkred symbol=circlefilled) ;
legenditem name='median' type=MARKER / label='Median' MARKERATTRS=(color=navy symbol=diamondfilled) ;
keylegend 'mean' 'median' /AUTOITEMSIZE;
keylegend 'box'/location=inside position=ne across=1;
run;
title;
%mend;

%box_plot(dsn=have,group=sex,x=bp_status,y=weight, title=VCSS-Socre)

[ballardw]

---

@JFuchsia wrote:

I tweaked the  previous scale statement in the current code to this :

yaxis integer values=(0 to 500 by 10);

 and the scale is much better and I don't lose my plot labels (for mean, min, etc.)

however, how can I group the data by each year?

Thank you!!

---

Not exactly clear what sort of "group" you are looking for. The Group=variable option on the plot? But since we don't know  your data or even the actual code you are running can't tell.

A separate graph for each year? That might be a BY statement. Or SGPANEL and PANELBY your year variable.

Or overlay multiple plots using different data layout.

[JFuchsia]

See new post. Thank you! 🙂