Date:         Sat, 12 Jan 2008 16:30:22 +0800
Reply-To:     "Hu, Jiangtang" <jiangtanghu@GMAIL.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         "Hu, Jiangtang" <jiangtanghu@GMAIL.COM>
Subject:      Re: cumulative sum with subgroups ?
Comments: To: Warren Schlechte <Warren.Schlechte@tpwd.state.tx.us>
In-Reply-To:  <43C07A163F7E764A8B27F6DAE2B126BF19BC32E6@tpwd-mx0.tpwd.state.tx.us>
Content-Type: text/plain; charset=ISO-8859-1

A modified version of Warren's, delete "retain cumsum", and replace
"cumsum=sumnb+cumsum" with " cumsum+sumnb":

data out;
    set test;
    by id;
    if first.id then cumsum=sumnb;
        else cumsum+sumnb;
run;

On Jan 12, 2008 1:15 AM, Warren Schlechte <Warren.Schlechte@tpwd.state.tx.us>
wrote:

> data test;
> input id      date    nnb     sumnb;
> datalines;
> 1       10      2       15
> 1       11      1       15
> 1       12      1       20
> 1       13      1       10
> 2       7       1       150
> 2       9       1       200
> 2       10      2       150
> 2       17      1       300
> ;
> run;
>
> proc sort data=test; by id;run;
>
> data out;
> set test;
> by id;
> retain cumsum;
> if first.id then cumsum=sumnb; else cumsum=sumnb+cumsum;
> run;
>
> proc fsview data=out;run;
>
> Warren Schlechte
>
>
> -----Original Message-----
> From: ChrisG [mailto:chris.godlewski@GMAIL.COM]
> Sent: Friday, January 11, 2008 10:51 AM
> Subject: cumulative sum with subgroups ?
>
> Hi
>
> Suppose i have such dataset :
>
> id      date    nnb     sumnb
> 1       10      2       15
> 1       11      1       15
> 1       12      1       20
> 1       13      1       10
> 2       7       1       150
> 2       9       1       200
> 2       10      2       150
> 2       17      1       300
>
> i would like sas to do the following thing in a additional variable,
> let call it cumsumnb :
>
> id      date    nnb     sumnb    cumsumnb
> 1       10      2       15             15
> 1       11      1       15             30
> 1       12      1       20             50
> 1       13      1       10             60
> 2       7       1       150            150
> 2       9       1       200             350
> 2       10      2       150            500
> 2       17      1       300            800
>
> in other words, to calculate in an 'incremental way' the cumulative
> sum of sumnb by date within the id subgroup ...
>
> i hope i was clear enough ...
>
> many thanks for any help !
>
> cheers
>
> CG
>



--
John Jiangtang Hu