topic Grouping data in order to count frequencies in SAS Programming

Grouping data in order to count frequencies

deleted_user — Mon, 29 Jun 2009 18:44:49 GMT

Hello,

I am relatively new to using SAS beyond the routine PROC proceedures.

I am trying to create a simple frequency distribution chart, but on grouped data.
In other words, I want to define 10 groups, say 0-10, 10-20, ..... 90-100
and then count how many of the originial observations are in each of these 10 groups.
Is there an easy way to do this using PROC UNIVARIATE or such?

Re: Grouping data in order to count frequencies

Doc_Duke — Mon, 29 Jun 2009 22:49:55 GMT

The easiest way is to create a FORMAT (see PROC FORMAT for syntax) and then use PROC FREQ to get your frequencies.

Doc Muhlbaier
Duke

Re: Grouping data in order to count frequencies

DanielSantos — Tue, 30 Jun 2009 08:23:39 GMT

Doc's suggestion (PROC FORMAT+PROC FREQ) is just right.

More solutions are available, but all very similar, and all within a two pass execution.

Another one, would be two SQL queries or one SQL query with sub-query, (which is still a two pass solution), with the inner query producing some temporary GROUP_ID for each row (say GROUP_ID=1 case when 0-10, GROUP_ID=2 case when 10-20, ...), and the outer query counting row for each GROUP_ID.

Cheers from Portugal.

Daniel Santos @ www.cgd.pt.

Re: Grouping data in order to count frequencies

deleted_user — Tue, 30 Jun 2009 12:30:41 GMT

I am surprised SAS has not incoporated an easier way to do this.

...... are you all enjoying you legal drugs in Protugal.... just kidding. Wouldn't make a difference to me, but it makes sense economically.

Re: Grouping data in order to count frequencies

deleted_user — Tue, 30 Jun 2009 12:37:24 GMT

.... does anybody know of an explicit (simple) example where this is done. I am not familiar with the SQL syntax or even the FORMAT syntax.

....... it's no big deal, but I would like to learn it in case I come across it again.

Re: Grouping data in order to count frequencies

DanielSantos — Tue, 30 Jun 2009 13:29:36 GMT

But, it is indeed an easy procedure.

For example:

proc sql;
select GROUP_ID, count(*) as COUNT from (
select int(VALUE/10) as GROUP_ID from INPUT
) group by GROUP_ID;
quit;

int(VALUE/10) will return the integer value of VALUE/10, so
0 =< VALUE < 10 => GROUP_ID = 0
10 =< VALUE < 20 => GROUP_ID = 1
etc...

And by the way, drugs are still pretty illegal here.
It was discussed some time ago, but it was a no go idea.

Cheers from Portugal.

Daniel Santos @ www.cgd.pt.

Re: Grouping data in order to count frequencies

deleted_user — Tue, 30 Jun 2009 20:58:10 GMT

Wow, that's awesome.

I thought I heard on NPR that they decriminalized all drugs in Protugal???