topic Re: q-q plot using SGPlot or SG procedures in Graphics Programming

q-q plot using SGPlot or SG procedures

AmitKB — Wed, 23 Sep 2009 16:27:21 GMT

Hi all,
I want to create a normal probability plot or qq plot using SG procedures.

I know I can do it using proc univariate, but is it possible using SG procedures.

Thank for all your help.

Regards,

Amit

Re: q-q plot using SGPlot or SG procedures

DanH_sas — Wed, 23 Sep 2009 17:27:17 GMT

Hey Amit,

ODS GRAPHICS is also supported in PROC UNIVARIATE. Is there a particular reason that you need to use the SGPLOT?

Thanks!

Re: q-q plot using SGPlot or SG procedures

AmitKB — Wed, 23 Sep 2009 17:33:18 GMT

Hi Dan,
I am trying to develop a report for presentation. The Histogram in SGplot looks more clean than in proc univariate.
Thats the reason I wanted to see if I replecate my q-q plot using proc sgplot.

Re: q-q plot using SGPlot or SG procedures

DanH_sas — Wed, 23 Sep 2009 17:58:59 GMT

Right now, we do not have a qqplot statement in SGPLOT. However, if you turn on ODS GRAPHICS for UNIVARIATE, you should get a comparable histogram to the one from SGPLOT. Give it a try and see what you think.

Thanks!
Dan

ods graphics on;
proc univariate data=sashelp.class;
histogram weight;
run;
ods graphics off;

proc sgplot data=sashelp.class;
histogram weight;
run;

Re: q-q plot using SGPlot or SG procedures

BarryDeCicco — Mon, 12 Dec 2016 15:23:16 GMT

I just ran the two histograms. The ODS graphics statement look a little better (better shading), but they were very similar.

Re: q-q plot using SGPlot or SG procedures

Rick_SAS — Mon, 12 Dec 2016 15:32:46 GMT

Yes, you can do it. You need to compute the quantiles first. If you want a normal Q-Q plot, use PROC RANK:

/* get sample mean and stddev */
proc means data=sashelp.class mean std;
var weight;
run;

proc rank data=sashelp.class normal=blom out=QQ;
   var weight;
   ranks Quantile;
run;
 
proc sgplot data=QQ noautolegend;
   scatter x=Quantile y=weight;
   lineparm x=0 y=100 slope=22.77; /* y-int=mean slope=stddev */
   xaxis label="Normal Quantiles";
run;

For more discussion, see

http://blogs.sas.com/content/iml/2016/11/23/sampling-variation-small-samples.html

and for the general case of Q-Q plots for arbitrary distributions, see

http://blogs.sas.com/content/iml/2011/10/28/modeling-the-distribution-of-data-create-a-qq-plot.html