Date: Tue, 19 Dec 2006 05:59:14 -0800
Reply-To: SK <skauchali@GMAIL.COM>
Sender: "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From: SK <skauchali@GMAIL.COM>
Organization: http://groups.google.com
Subject: Re: Algorithm to create new var based on knowledge of date and
In-Reply-To: <200612191119.kBJ6Oefs016021@mailgw.cc.uga.edu>
Content-Type: text/plain; charset="iso-8859-1"
Yes, thanks, with this method, we have identified all the results that
have a yes and date then it was yes-- something I wanted. Also we have
marked those that did not have yes in the series of tests as no. Also
what I wanted.
because I am doing an analysis that aims to find out the exact time
when the result were YES, I need to know the last time when it was NO
and the first time it was YES. So the final dataset should look like
this:
childid date result
datefirstyes datelastno
test
101691 02JUL2002 No
101691 13AUG2002 Yes
13AUG2002 02JUL2002 YES
101691 01OCT2002 Yes
101691 05NOV2002 Yes
101691 26AUG2003 Yes
101691 30SEP2003 Yes
In this case I would be able to say that the test became YES somewhere
between <datelastno> to <datefirstyes>. Those subjects that did not
have any YES in the follow-up would only have a <datelastno> at the
last DATE, i.e. datefirstyes would be missing for this child. For those
that start the follow-up time with a YES, will have datefirstyes but no
datelastno.
Thanks for the help, it is indeed much appreciated.
SK
On Dec 19, 1:19 pm, gerhard.hellrie...@T-ONLINE.DE (Gerhard Hellriegel)
wrote:
> so you need a dataset with all children fed up with the information when the
> test was positive the first time, or missing if there was none. Is that right?
> Have a look at the following:
>
> proc sort data=children;
> by childid date;
> run;
>
> data children;
> retain res;
> set children;
> by childid;
> if first.childid then res=0;
> if upcase(result)="YES" then res+1;
> run;
>
> proc sort;
> by childid res;
> run;
>
> data children;
> set children;
> retain test;
> by childid;
> if res=0 then test=.;
> if res=1 then test=date;
> if last.childid then output;
> format test date9.;
> run;
>
> Is that what you need?
> Regards,
> Gerhard
>
> On Tue, 19 Dec 2006 02:36:37 -0800, SK <skauch...@GMAIL.COM> wrote:
> >Hi There (sorry this is a long message; I have tried to keep it clear;
> >let me know if there is further information you require)
>
> >Let me start by admitting that I am not an expert at repeated
> >observation in SAS, and I think I should spend time learning it before
> >I resort to asking more question here. I have Cody's book (long. data
> >and SAS), which seems to help. But I thought I would post to fast track
> >my objectives. Here is the problem: data table is as
>
> >Childid date result
> >100611 19FEB2002 No
> >100611 11JUN2002 No
> >100611 08JUL2003 No
> >100611 08JUL2004 No
> >100651 09APR2002 No
> >100651 15OCT2003 No
> >101061 30APR2002 No
> >101061 27AUG2002 No
> >101061 12NOV2002 No
> >101121 16MAY2002 No
> >101121 16JUL2002 Yes
> >101161 22JAN2002 No
> >101161 21MAY2002 No
> >101271 11DEC2001 No
> >101271 02APR2002 No
> >101271 22APR2003 No
> >101361 02APR2002 No
> >101361 25JUL2002 No
> >101361 19MAY2003 No
> >101371 29JAN2002 No
> >101371 21MAY2002 No
> >101371 01JUL2003 No
> >101391 26MAR2002 No
> >101391 23JUL2002 No
> >101391 19MAY2003 Yes
> >101561 26MAR2002 No
> >101611 02JAN2002 No
> >101611 29JAN2002 No
> >101651 26MAR2002 No
> >101651 24JUL2002 No
> >101651 01SEP2003 No
> >101691 02JUL2002 No
> >101691 13AUG2002 Yes
> >101691 01OCT2002 Yes
> >101691 05NOV2002 Yes
> >101691 26AUG2003 Yes
> >101691 30SEP2003 Yes
> >101711 21MAY2002 Yes
> >101711 02JUL2002 Yes
> >101711 03SEP2002 Yes
> >101711 14JUL2003 Yes
> >307511 30NOV2004 Yes
> >307601 26APR2005 No
>
> >As you can see, a child has had either just one measurement or multiple
> >measurements (result) at different dates. The RESULT is a test result.
> >If the child test YES at a date, then the test is positive
> >(TEST=positive). If RESULT is NO on all dates then the TEST=negative.
> >What is important is the date when the RESULT was YES.
>
> >I would like to program in SAS an algorithm that will identify a child
> >who is test positive and date when it was positive. So for example,
> >these 2 children
> >307511 30NOV2004 Yes
> >307601 26APR2005 No
> >only had one test done. If child only has one obs then TEST=RESULT and
> >DATETEST=DATE- in these 2 cases the data will look like this:
>
> >Childid date result test datetest
> >307511 30NOV2004 Yes Yes 30NOV2004
> >307601 26APR2005 No No 26APR2005
>
> >In cases where the child had more than one obs, it is a bit tricky for
> >me. If any of the dates had a RESULT=YES, then TEST=YES and DATETEST=
> ><date when first time when RESULT=YES>; Take this subject as an
> >example:
>
> >Childid date result test
> >datetest
> >101691 02JUL2002 No
> >101691 13AUG2002 Yes Yes 13AUG2002
> >101691 01OCT2002 Yes
> >101691 05NOV2002 Yes
> >101691 26AUG2003 Yes
> >101691 30SEP2003 Yes
>
> >If all of the results on all dates are NO, then the test is also NO;
> >and the last date of result is recorded; thus:
> >Childid date result test
> >datetest
> >101361 02APR2002 No
> >101361 25JUL2002 No
> >101361 19MAY2003 No No 19MAY2003
>
> >If I look at this, it would seem that subjects with only one
> >measurement are a special case of multiple obs (where the wave is equal
> >to one).
>
> >I would like to then take the obs that have a test and datetest var
> >populated as above from this dataset and merge it with the flat data
> >set with all children in it. Can someone help me with the code and
> >direction.
>
> >Thanks
> >SK
| 
