See distributions of student height by grade level

[NKormanik]

Would someone please steer me in the right direction for creating the following:

In the example above, there are four grade levels, and height generally overlaps to a degree.

Ideally the plot should show a distribution, clustering toward the center of each grade, and outliers at either end.

Thanks very much!

Nicholas Kormanik

[RW9]

My goto place for anything graph related is this site:

http://blogs.sas.com/content/graphicallyspeaking/

There are thousands of examples there.

[PeterClemmensen]

Follow this example and use HBOX instead of VBOX in PROC SGPLOT.

http://blogs.sas.com/content/graphicallyspeaking/2017/06/16/scatter-mean-value/

Plot sample data if you want a code answer 🙂

[stat_sas]

Hi,

You can try proc boxplot for this type of analysis. It will compare distribution of height at each grade. Also highlights outliers and skewness.

[Rick_SAS · Accepted solution]

I like draycut's solution because it seems closest to the image that you posted. However, be aware that box plots show a schematic representation of the distribution, and jittering breaks down when you get to thousands of observations. For larger samples, look at comparative histograms, which scale to larger samples.

[NKormanik]

Rick, I think the first plot from your article will work well for my purpose.

Code:

proc univariate data=sas_1.divisions_20905;
  class Rank;
  var i_20905;
  histogram i_20905 / nrows=7 odstitle="i_20905";
  ods select histogram;
run;

Two follow-up questions:

1. Could we overlay some statistical information within the plot, such as percentile numbers?

2. Output plot files need to be appropriately named.  In the case above, "i_20905".

Thanks so much for your help.  As well as to the others here.

[NKormanik]

The following seems to work out pretty well.  Thanks again for everyones help.

ods graphics on / reset=index imagename="20905";

proc univariate data=sas_1.divisions_20905;
class rank (order=data);
var i_20905;
histogram i_20905 / nrows=7 odstitle="20905";
inset nobs max p95 p75 mean p50 p25 p5 min / format=6.1 pos=nw;
ods select histogram;
run;

[ballardw]

Doesn't show "outliers" but something like this perhaps:

data example;
   do grade= 1 to 4;
      do i=1 to 1000;
         height = grade*2 + round(rand('uniform')*36);
         output;
      end;
   end;
run;

proc sgplot data=example;
   density height /group=grade;
run;