Superimpose a graph

[deleted_user]

How can I superimpose a scatterplot to a boxplot?

[DanH_sas]

Assuming you have SAS 9.2, you can do this easily using the Graph Template Language. Here is a simple example:

proc template;
define statgraph boxscatter;
begingraph;
layout overlay;
boxplot x=age y=height;
scatterplot x=age y=weight / yaxis=y2;
endlayout;
endgraph;
end;
run;

proc sgrender data=sashelp.class template=boxscatter; run;

[deleted_user]

Thanks, for the code. Do I have problem with this code if the data sources for the plots are not same. I also need to include a by variable in the boxplot. In addition, I got the following message in the log:

1143 proc template;
1144 define statgraph boxscatter;
1145 begingraph;
----------
180
ERROR 180-322: Statement is not valid or it is used out of proper order.
1146 layout overlay;
1147 boxplot x=age y=height;
1148 scatterplot x=age y=weight / yaxis=y2;
1149 endlayout;
1150 endgraph;
--------
1
WARNING 1-322: Assuming the symbol END was misspelled as endgraph.
1151 end;
WARNING: Object will not be saved.
1152 run;
NOTE: PROCEDURE TEMPLATE used (Total process time):
real time 0.21 seconds
cpu time 0.00 seconds

WARNING: Errors were produced.
NOTE: The SAS System stopped processing this step because of errors.
1153


1154 proc sgrender data=sashelp.class template=boxscatter; run;

ERROR: Procedure SGRENDER not found.

NOTE: The SAS System stopped processing this step because of errors.
NOTE: PROCEDURE SGRENDER used (Total process time):
real time 0.00 seconds
cpu time 0.00 seconds

[GraphGuy]

Here's how with good-old traditional gplot (the example generates some fake/random data, and also plots it)...

Assuming you're using the same data for the boxplot and the scatter plot (ie, you just want to also show a scatter plot of the points that the boxplot is based on), this can be very easily done by putting the y*x on the plot statement twice, and using a different plot interpolation for each:

data foo;
do x=0 to 10 by 1;
do i=1 to 20;
y=rannor(23749);
output;
end;
end;
run;

symbol1 v=none i=box color=blue;
symbol2 v=dot height=.01 i=none color=red;
proc gplot data=foo;
plot y*x=1 y*x=2 / overlay;
run;



Or, you can plot totally separate data (that isn't necessarily confined to having the exact same x-values as the boxplot), using a technique such as this...

data foo;
do x=0 to 10 by .1;
do i=1 to 20;
y=rannor(23749);
output;
end;
end;
run;

data foo; set foo;
x_rounded=round(x);
run;

symbol1 v=none i=box color=blue;
symbol2 v=dot height=.01 i=none color=red;
proc gplot data=foo;
plot y*x_rounded=1 y*x=2 / overlay;
run;

[deleted_user]

This code works. Thanks a lot.

[DanH_sas]

Based on your log output, you must not have SAS 9.2. Robert's gplot approach should work fine for you. As a side note, the SGRENDER procedure does support BY-groups. The BY-group variable(s) is not specified in the template.

Thanks!
Dan

[DanH_sas]

Correction on my last post, you probably have 9.2 phase 1. Does the code run for you if you remove the BEGINGRAPH and ENDGRAPH?

Thanks!
Dan

[DanH_sas]

In addition, you may need to use this bit of code instead of SGRENDER:

data _null_;
set sashelp.class;
file print ods=(template="boxscatter");
put _ods_;
run;