Date:         Mon, 8 Sep 2008 08:31:38 -0700
Reply-To:     snoopy369@GMAIL.COM
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         snoopy369@GMAIL.COM
Organization: http://groups.google.com
Subject:      Re: Alternate for retain
Comments: To: sas-l@uga.edu
Content-Type: text/plain; charset=ISO-8859-1

On Sep 8, 10:17 am, pausha <paus...@gmail.com> wrote:
> i have this dataset its clinical
>
> which has like 100 week visit and it slike for every 8 weeks my nee
> dis i have to put  a flag for subject who have switched in dose after
> week 52.
> the scenario is they could have switched dose even before week 52 but
> i have to consider only afetr week 52
> dataset is like
> sub   visit            dose
> 001    week 50     30mg
> 001    week 51     60mg
> 001    week 76     60mg
> 001    week 80     60mg
>
> in this case his flag should be 'No' because his change is not after
> week 52
>
> sub  visit              dose
> 002  week 35       30mg
> 002  week 52       60mg
> 002  week 58       90mg
>
> in this case his flag is 'Yes' because his dose had incresed from 60
> to 90 after week 52
>
> each subjec have like 10 records so i need to use anythinng other than
> reatin and first.sub
>
> any different ideas appreciated for this  look check and flagging.
> Thanks

Sounds like the easiest way to do this is (assuming 'visits' is the
dataset above):

proc sort data=visits; by sub visit dose; run;
data flags;
set visits;
by sub visit dose;
if first.dose /* and not first.sub */ and visit > 52 then flag =
"yes";  *this assumes visit is a numeric field, you need to use substr
or whatever if it actually contains "week 52";
run;
with optionally a check for first.sub if you want to not flag people
whose first visit was after week 52 (if that's possible in your
study), as they'd flag on their first dose without that check.  I
don't think a retain would be necessary.