| Date: | Thu, 20 Nov 2003 10:44:40 -0500 |
| Reply-To: | diskin.dennis@KENDLE.COM |
| Sender: | "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU> |
| From: | Dennis Diskin <diskin.dennis@KENDLE.COM> |
| Subject: | Re: How To Match/Merge Multiple Observations to Multiple
Observations. |
|
| Content-Type: | text/plain; charset="us-ascii" |
|---|
Pat,
Harry Droogendyk has already given you a fairly standard approach to this type of problem.
Given the large number of SMF records and the relatively small number of
LU, I would suggest using a format by preprocessing the LUNAME dataset.
You would have to verify that there is no overlap in dates for a given LU
(probably needed anyway) and than you could produce a format with ranges
comprised of concatenated LUid and date (where the date must be formatted
as yyyymmdd) such as:
'LU12319990923'-'LU12320001024'='PrinterA'
Be sure to include an OTHER entry to account for non-matches. e.g.
other='Unknown'
This can all be done via datasteps and proc format using the CNTLIN
option, not manual coding.
In your SMF processing you would then build the LU concatenated with the
date and apply the format to get a printer name.
Is this clear ? I can get more detailled if you need it.
Regards,
Dennis Diskin
Pat Monk <Pat_Monk@US.CRAWCO.COM>
Sent by: "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
11/20/2003 09:51 AM
Please respond to Pat_Monk
To: SAS-L@LISTSERV.UGA.EDU
cc:
Subject: How To Match/Merge Multiple Observations to Multiple Observations.
I have a SAS database containing 4MM+ observations built from SMF type6
records. I have a second "control" database that I need to "match" to the
first database on a variable named LUNAME.
The control database contains observations that identify the remote
printer id (Rnnnn) for a given LUNAME as of a certain date range. Now for
the problem...
The control database may contain multiple observations for a given LUNAME,
based upon a START DATE-END DATE range identified in the control record.
In other words, for a given LUNAME found in an SMF type6 record, the
associated Rnnnn is to be determined by the date within the type6 SMF
record falling in a given date range in a matching control record.
I have not found a technique that will successfully match the SMF records
to the appropriate control record in order to determine the correct Rnnnn
for any given date.
Can you suggest a solution to my multiple-to-multiple match problem?
Pat Monk
Senior Network Systems Programmer
IT Technical Services
Office: 404 843-6203
Fax: 404 705-6817
|