Date: Fri, 17 Dec 2004 16:26:44 -0500
Reply-To: Nathaniel_Wooding@DOM.COM
Sender: "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From: Nat Wooding <Nathaniel_Wooding@DOM.COM>
Subject: Re: Complicated aggregate function?
Content-type: text/plain; charset=US-ASCII
Hjalte
This is really not a hard problem. There should be several solutions but
here is mine. As you will see, the problem will need to be split into
several steps.
Nat Wooding
data a;
input col1-col3;
cards;
1 2 3
4 5 6
7 8 9
10 11 12
13 14 15
16 17 18
19 20 21
22 23 24
data b; * FIRST IDENTIFY THE GROUPS . THE MOD FUNCTION WILL NEED TO BE
CHANGED;
set a;
if _n_=1 then do;
group=1;
count=0;
end;
if mod(count,2)=0 then group+1;* THE DIVISOR (2) SHOULD BE 15000 for
your work;
count+1;
proc means noprint;
var col:;
OUTPUT OUT=B SUM= ;
BY GROUP; * AGGREGATE OVER GROUPS;
PROC TRANSPOSE OUT=B;* NOW SWAP ROWS AND COLUMNS;
VAR COL:;
ID GROUP;
PROC PRINT;
RUN;
Hjalte
<damhj@YAHOO.DK> To: SAS-L@LISTSERV.UGA.EDU
Sent by: "SAS(r) cc:
Discussion" Subject: Complicated aggregate function?
<SAS-L@LISTSERV.U
GA.EDU>
12/17/04 03:55 PM
Please respond to
Hjalte
Hi,
I would really appreciate some help with the following problem:
I have a SAS dataset with 3 variables and 150000 rows. I now want to
get the sum of each of the three variables, respectively. The problem
is that I need to group them by decentiles, that is: I want it to make
a group out of the first 1-15000 rows, a group out of the next
15001-30000 rows, and so on, until I have 10 groups with 15000 rows in
each, i.e. a table with 10 rows and three columns displaying the
corresponding sum in each. Is there an easy way to achieve this? I
would be most grateful for any help!
Best wishes,
Hjalte Dam, student at the M.Sc. in IT, communication and organization
at the University of Southern Denmark
