topic Re: Proc univariate for distribution histogram in SAS Procedures

Proc univariate for distribution histogram

DmytroYermak — Mon, 17 Sep 2018 11:52:06 GMT

Hi all,

I have the following training data:

data test;
  infile datalines dlm="," dsd;
  input severity @@;
datalines;                      
0,0,0,1,1,1,1,1,1,1,1,1,1,1,2,2,2,2,3,3,3,3,4,4,5,5,5,5,6,7,9,10,11
;
run;

How to built a distribution histogram with visualisation of Median and percentiles using proc univariate.

Thank you!

Re: Proc univariate for distribution histogram

PaigeMiller — Mon, 17 Sep 2018 12:34:30 GMT

If it has to be done in PROC UNIVARIATE, I think you can't do that. If you want to do this in PROC UNIVARIATE followed by PROC SGPLOT, then it's pretty simple.

Re: Proc univariate for distribution histogram

Reeza — Mon, 17 Sep 2018 14:54:34 GMT

How would you want the graph to look? AFAIK, that isn't a standard histogram and how you identify the percentiles and median can vary.

Re: Proc univariate for distribution histogram

Rick_SAS — Tue, 18 Sep 2018 00:31:59 GMT

You didn't say which percentiles you want, but her is how to do it with the 10th and 90th percentiles:

proc means data=test median p10 p90;
var severity;
run;

proc univariate data=test;
var severity;
histogram severity / href=(2.0 1.0 7.0);
inset P10 median P90 / position=NE;
run;

If you prefer the 5th and 95th percentiles, use P5 and P95 in the program.

Re: Proc univariate for distribution histogram

DmytroYermak — Thu, 20 Sep 2018 13:02:07 GMT

Thank you, Rick. I have added the "normal" option to see the normal plot approximation(sorry for "my terminology"). I have just one question: how to output "zero" values on the histogram as well

Re: Proc univariate for distribution histogram

Rick_SAS — Thu, 20 Sep 2018 13:26:36 GMT

Your histogram is already displaying zero values. The first bin displays the counts of all values in the half-open interval [0, 2).

That fact might be clearer if you use the ENDPOINTS option to display the endpoints of the bins instead of the midpoints:

histogram severity / href=(2.0 1.0 7.0) endpoints;

You can also change the default bin widths and locations, if necessary.

Re: Proc univariate for distribution histogram

DmytroYermak — Thu, 20 Sep 2018 14:16:27 GMT

Well, how to set up the same interval that we have in our dataset [0,1)?

Re: Proc univariate for distribution histogram

Rick_SAS — Thu, 20 Sep 2018 14:20:43 GMT

The last line in my previous message links to an article that explains everything.

Re: Proc univariate for distribution histogram

DmytroYermak — Thu, 20 Sep 2018 14:22:05 GMT

Thank you! That is what I was looking for!