topic Marking outlier names on graph in Graphics Programming

Marking outlier names on graph

ieva — Wed, 02 Nov 2011 08:01:49 GMT

Hi,

I would like to make a graph by 2 variables, create different colour labels for 3 different groups and be able to marke some specific observations (like outliers) in final graph.

So my code:

proc gplot data=mydata;

plot var2*var1=groupname/frame cframe=ligr

legend=legend1 vaxis=axis1 haxis=axis2;

run;

What I get what I need - values of those 2 variables by group names, but I can’t understand is there any possibility to mark in the graph which are those outlying observations?

I would appreciate if somebody of you could help.

Thank you,

Ieva

Marking outlier names on graph

GraphGuy — Wed, 02 Nov 2011 13:27:02 GMT

One way to selectively label certain outliers is to selectively assign text to a variable,

and then use that variable as the 'pointlabel' (assuming you're using a fairly recent

release of SAS)...

data foo; set sashelp.class;
if height>=67 then outlier_text=trim(left(height));
run;

symbol1 value=dot interpol=none pointlabel=("#outlier_text");

proc gplot data=foo;
plot height*weight;
run;

Marking outlier names on graph

ieva — Thu, 03 Nov 2011 08:37:38 GMT

Thank you for suggestion!

I'm using SAS 9.1.3. When running your code I get labels for all observations. And when I try to change outlier_text to variable name (not height) also nothing is changing. Is my SAS version too old for this?

Marking outlier names on graph

GraphGuy — Thu, 03 Nov 2011 12:07:01 GMT

SAS 9.1.3 is over 7 years old but I tried both of those things in it just now, and they both work for me.

Is it possible you're using device=java or device=activex ?

They only have "partial support" of a lot of the features like pointlabel.

Marking outlier names on graph

ieva — Thu, 03 Nov 2011 12:14:08 GMT

Ah, then it is because device=activex.

Marking outlier names on graph

Rick_SAS — Fri, 11 Nov 2011 15:36:00 GMT

You can also mark outliers in a scatter plot by using ODS statistical graphics (the SGPLOT procedure). See http://blogs.sas.com/content/iml/2011/11/11/label-only-certain-observations-with-proc-sgplot/

Marking outlier names on graph

ieva — Mon, 14 Nov 2011 07:11:51 GMT

Thank you! This seems as a good, simple aproach!

Marking outlier names on graph

Alfredo — Wed, 18 Jan 2012 12:11:19 GMT

If you want to have more control over your graph, try the annotate facility:

proc sql;

create table anno as select '2' as xsys, '2' as ysys, height as y, weight as x,

name as text, 'label' as function,'"Arial"' as style, 'red' as color, '3' as position

from sashelp.class

having height not between avg(height)-2*std(height) and avg(height)+2*std(height) ;

quit;

ods html body='c:\temp\myGraph.html';

goptions dev=activex;

proc gplot data=sashelp.class;

plot height*weight / anno=anno;

run;quit;

ods html close;