topic Re: Group information in SAS Procedures

Group information

deleted_user — Wed, 27 May 2009 20:22:41 GMT

I have information of selling by date and I've already have the week number for each observation, now I need to consolidate it by week

Table content

Date.............Client......Sku......Week.....Volume
01/10/08.......3010......9999....40.............10
02/10/08...... 3010......9999....40.............05
03/10/08...... 3010......9999....40.............30

01/10/08.......3010......8888....40.............20
02/10/08...... 3010......8888....40.............10
03/10/08...... 3010......8888....40.............60

01/10/08.......4000......5555....40............02
02/10/08...... 4000......5555....40............01
02/10/08...... 4000......5555....40...........01

01/10/08.......4000......3333....40............10
02/10/08...... 4000......3333....40............01
02/10/08...... 4000......3333....40...........01

Output expected

Client.....Sku.....Week.....Volume
3010......9999.....40.........45
3010......8888....40..........80
4000......5555.....40....... 04
4000.....3333......40...... 12

Thanks

Re: Group information

sbb — Wed, 27 May 2009 20:39:30 GMT

Investigate using SAS PROC SUMMARY.

Scott Barry
SBBWorks, Inc.

SAS 9.2 DOC: The SUMMARY Procedure
http://support.sas.com/documentation/cdl/en/proc/61895/HTML/default/a000146755.htm

Re: Group information

deleted_user — Wed, 27 May 2009 20:48:19 GMT

Hope the following the code helps:

proc sql;
create table test2 as
select client, week, sku, sum(volume) as volume
from testdata
group by sku
quit;

proc sort data=test2;
by sku;
run;

data test2;
set test2;
by sku;
if first.sku;
run;

~ Sukanya E

Re: Group information

Cynthia_sas — Wed, 27 May 2009 20:56:49 GMT

Hi:
As Scott says, either PROC MEANS or PROC SUMMARY can summarize the data by any combination of CLASS variables.

However, I'm confused by the data. I don't understand how Jan 10, Feb 10 and March 10 are all week number 40. Unless the WEEK variable doesn't have any relevance to the actual DATE variable.

I may have misinterpreted your starting data and your desired results, but it seems to me as though WEEK is irrelevant in this context and the desired report is actually a summary based on CLIENT and SKU.

cynthia

Re: Group information

deleted_user — Wed, 27 May 2009 22:03:27 GMT

I'm Sorry about the confusion. Data format is DD/MM/YY. In fact week 40 is october 1st, 2nd and 3th

I need to summarize by SKU, client per week
Tks Message was edited by: elilika

Re: Group information

deleted_user — Wed, 27 May 2009 22:19:05 GMT

The code provided earlier was summerized by SKU, and the output was what you expected. Let me know if that is not the desired result.

~ Sukanya E Message was edited by: Sukanya

Re: Group information

ChrisNZ — Thu, 28 May 2009 02:47:14 GMT

As Scott said, proc SUMMARY does it:

[pre]class CLIENT WEEK SKU;
var VOLUME;
output out=DSOUT sum=;