|
New Text Item: Re[2]: comm1-comm6 problem
It must be Friday afternoon. No sooner did I post the following reply
than it occurred to me that I probably misunderstood what Lisa meant
when she stated:
">I have a total of about 67 codes (numeric), any of which can be found
>among six consecutive fields (comm1 comm6), although for each
>observation, a code can only show up once."
At first I took this to mean that all but one of the the variables
comm1-comm6 would have missing values, leaving only one of the variables
containing a code. (see my first reply)
A more reasonable interpretation is that a code will not repeat itself
in a observation, but comm1-comm6 will all have different codes. If
that is the case, then this code should produce the desired report. (if
you want an output dataset named "COMMOUT". then put "create table
commout as" before the select statement):.
proc sql;
select key, count(*)
from ( select comm1 as key from comfile
union all (select comm2 as key from comfile)
union all (select comm3 as key from comfile)
union all (select comm4 as key from comfile)
union all (select comm5 as key from comfile)
union all (select comm6 as key from comfile))
group by key;
The union statement can be a bit tricky. Be careful when you use it!
______________________________ Reply Separator ____________________________
_____
Subject: Re[2]: comm1-comm6 problem
Author:
Dave_Mabey_at_RDA8POSTOFFICE1%RDCCMAIL%READERSDIGEST@RMAIL.COM@INTERNET
at RDNOTES
Date: 1/17/97 1:36 PM
(Embedded image moved to file: PIC001.PCX)
New Text Item: Re: comm1-comm6 problem
Lisa,
I would use the following.
PROC SQL;
select case when comm1 ne . then comm1
when comm2 ne . then comm2
when comm3 ne . then comm3
when comm4 ne . then comm4
when comm5 ne . then comm5
when comm6 ne . then comm6
when comm7 ne . then comm7
else -1
end as index,
count(*)
from comfile
group by index;
** Generic Signature **
______________________________ Reply Separator ____________________________
_____
Subject: Re: comm1-comm6 problem
Author: j_weedon@ESCAPE.COM@INTERNET at RDNOTES
Date: 1/16/97 11:11 AM
(Embedded image moved to file: PIC001.PCX)
On Thu, 16 Jan 1997 08:51:23 -0500, Lisa Bastian <eagle89@EROLS.COM> wrote:
>I have a total of about 67 codes (numeric), any of which can be found
>among six consecutive fields (comm1 comm6), although for each
>observation, a code can only show up once. Is there any "easy" way to
>count the occurrence of each individual code? ie., without an array and
>a bunch of "if" statements? Thanx.
Please put a subject line in your posts.
Does this do what you want:
%macro comms;
data new (keep=comm);
set old (keep=comm1-comm6);
%do i=1 %to 6;
comm=comm&i; output;
%end;
proc freq data=new;
run;
%mend; %comms
I'm not sure what your aversion to arrays is, but a non-macro solution with
an array would be:
data new (keep=comm);
set old (keep=comm1-comm6);
array comms comm1-comm6;
do over comms;
comm=comms; output;
end;
proc freq data=new;
run;
Jay Weedon.
SMTPOriginator: owner-sas-l@VM.MARIST.EDU
SMTPOriginator: owner-sas-l@VM.MARIST.EDU
| 
