Date:         Fri, 27 Aug 2010 15:12:18 -0400
Reply-To:     Michael Vostrich <mvostrich@MAIL.RU>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Michael Vostrich <mvostrich@MAIL.RU>
Subject:      How to recognize a pattern and assign a group
Content-Type: text/plain; charset=ISO-8859-1

Dear SAS-L experts,

I would greatly appreciate you help in finding a solution to a problem of
grouping the observations in the following situation. I have tests that were
performed in several groups, but the group number is missing. I know that
the tests in a group were performed on “seemingly consecutive” days, while
groups were separated by several weeks (at least, this is true within the
same test number). I need to group the records by assigning the group number
to a group of tests for the same test number.
The example dataset is dataset A below. I calculated the duration, which
gives me a very visual pattern (see the proc print), and I will appreciate
your help with the code that would create GROUPID, which I just added
manually myself to explain what I need (GROUPID serves to group the "similar
tests" within the same test number).

Thank you in advance!
Michael

data a;
format testdate date.;
  input testno testdate date.;
cards;
1  15mar2010
1  16mar2010
1  17mar2010
1  31mar2010
1  01apr2010
1  02apr2010
1  20apr2010
1  22apr2010
2  13mar2010
2  15mar2010
2  30mar2010
2  31mar2010
2  01apr2010
2  02apr2010
2  20apr2010
2  24apr2010
;

data b (drop=_:);
  set a ;
  by testno testdate;
  _temp=lag(testdate);
  if first.testno then _temp=.;
  duration=testdate - _temp;
run;
proc print data=b; run;

testdate    testno    duration      GROUPID
                                    manually added

15MAR02        1          .         1
16MAR02        1          1         1
17MAR02        1          1         1
31MAR02        1         14         2
01APR02        1          1         2
02APR02        1          1         2
20APR02        1         18         3
22APR02        1          2         3
13MAR02        2          .         1
15MAR02        2          2         1
30MAR02        2         15         2
31MAR02        2          1         2
01APR02        2          1         2
02APR02        2          1         2
20APR02        2         18         3
24APR02        2          4         3