Date:         Thu, 20 Nov 2003 14:45:41 GMT
Reply-To:     LWn <villerwalle.nononospam@YAHOO.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         LWn <villerwalle.nononospam@YAHOO.COM>
Organization: Telia Internet
Subject:      Re: Calculating time to an event

Sean

I don't have a solution but at least an idea for a start.
Since it looks like every weekvalue is a multiple of 26 I would start using
Proc Transpose to get all values from each ID in the same record.

ID  Score26  Score52  Score78  Score104   etc
1     20           20            17            .
2     15           .               18            .
3     12           .               16           15

thereafter a number of IF/THEN could do the rest.
hope it helps a bit

/LWn


"Candrilli, Sean D." <scandrilli@RTI.ORG> skrev i meddelandet
news:085D75F957E029479D75AD90F8A8F14B05822B@rtints3.rti.org...
> Greetings Fellow SAS Users,
>
> I am new to the group, and have a question that I am hoping someone can
> assist me with.  I need to estimate a proportional hazards model, which is
> the easy part.  The difficulty lies in the data that I received.  I having
a
> terrible time trying to get the data setup like it needs to be.
>
> I have something that looks like this:
> ID  Week  Score
> 1    26    20
> 1    52    20
> 1    78    17
> 2    26    15
> 2    78    18
> 3    26    12
> 3    78    16
> 3    104   15
>
> My event of interest is a drop in score, and I need to model the DAYS
> (notice that I don't have days here, but rather week of visit) to a drop
in
> score.  To further complicate things, I need to do this for a 1-point
drop,
> a 2-point drop, etc, etc.  With regards to censoring, for a censored term,
> the time to the event would be the last day seen (which may differ across
> respondents.)  Unfortunately I have respondents that skipped visits (ID #3
> above did not come in for the week 52 visit.)
>
> For ID #1 above I think that the days to events should be as follows (each
> 26 week block is set to 182 days):
> days_1pt_drop = 182 + 182 + 60.7 = 424.7
> days_2pt_drop = 182 + 182 + 60.7 + 60.7 = 485.4
> days_3pt_drop = 182 + 182 + 60.7 + 60.7 + 60.7 = 546
>
> and
> cens_1pt_drop = 0
> cens_2pt_drop = 0
> cens_3pt_drop = 0
>
> because clearly these events did happen for this ID.
>
> Now, for ID #2, who did not experience a decline in score, days should be
as
> follows:
> days_1pt_drop = 546
> days_2pt_drop = 546
> days_3pt_drop = 546
>
> because the last day this respondent was seen was on day 546 (78 weeks X 7
> days)
>
> and
> cens_1pt_drop = 1
> cens_2pt_drop = 1
> cens_3pt_drop = 1
>
> Ultimately I think that I need a file that looks like this:
>
> ID  days_1pt days_2pt days_3pt cens_1pt cens_2pt cens_3pt
> 1    424.7    485.4      546      0        0         0
> 2    546      546        546      1        1         1
>
>
> I apologize for dumping all of this here.  I really do not have any idea
how
> to go about attacking this problem.  If this isn't the appropriate forum
for
> something like this, and someone knows a more appropriate place to post
> this, I would appreciate learning of it.
>
> Any thoughts or suggestions would be greatly appreciated.
>
> Kind Regards,
>
> Sean