| Date: | Tue, 6 Oct 2009 14:33:49 -0400 |
|---|
| Reply-To: | Rieza Soelaeman <rsoelaeman@GMAIL.COM> |
|---|
| Sender: | "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU> |
|---|
| From: | Rieza Soelaeman <rsoelaeman@GMAIL.COM> |
|---|
| Subject: | Doing arithmetical operations on observations that meet certain
conditions for records w/ multiple observations to a unique
identifier... |
|---|
| Content-Type: | text/plain; charset=ISO-8859-1 |
|---|
Dear all,
Supposing I have data as follows:
personid specimen_type coll_dt count 001 A 1/1/2009 1 001 B 4/4/2009 2
001 C 4/4/2009 3 002 A 3/6/2009 1 002 B 6/21/2009 2 003 D 5/9/2009 1 003 A
5/30/2009 2 003 B 7/14/2009 3
I want to have SAS create a variable "testlag" that subtracts the coll_dt
(where specimen_type = B) from the coll_dt (where specimen_type=A). I
envision the final dataset to be something like this:
personid specimen_type coll_dt count testlag 001 A 1/1/2009 1 93 001 B
4/4/2009 2 001 C 4/4/2009 3 002 A 3/6/2009 1 107 002 B 6/21/2009 2 003 D
5/9/2009 1 003 A 5/30/2009 2 45 003 B 7/14/2009 3
Is it possible to do this directly using this "long" format data or would I
need to convert the data to wide first (using proc transpose or arrays), and
then subtract colldt1 from colldt 2.
Specifically, can I do
testlag = coll_dt (where specimen_type = 'B' and count = 2) - coll_dt
(where specimen_type=A and count = 1);, or, like I said above, will I need
to make it wide and do "testlag = colldt2 - colldt 1;"?
Thanks,
Rieza
--
RHS
| 
