Addressed to: "Lund, Peter" <Peter.Lund@OFM.WA.GOV> SAS-L@UGA.CC.UGA.EDU

** Reply to note from "Lund, Peter" <Peter.Lund@OFM.WA.GOV> Tue, 16 Sep 1997 16:35:04 -0700 > > I had a call from a colleague in another agency who has the following > situation: > > They contracted to have some data prepared - the contractor is using > PowerBuilder and a SQL Anywhere database for the work. The tables > range from 100,000 to 2,500,000 records. The contractor will also > provide software, written in PowerBuilder, to capture and load future > data. > > Here's the question. They want to do any analysis of the data with SAS > and wonder if the data should be left in SQL Anywhere and accessed via > ODBC or converted to SAS. If it weren't for the fact that future > (quarterly) data would be loaded into the database I would say to > convert it to SAS. I have a lot of experience using SAS and ODBC but > not against data tables as large as these. > > Any ideas are welcome - I'll pass them along. Another issue that is > potentially relevant - as far as I know, they have moderate (at best) > SAS experience and probably no database experience. I don't think they > have SQL Anywhere, but could probably get it. It's not the most user > friendly database, but it's workable. > > Thanks- > Pete Lund > WA State Office of Financial Management > peter.lund@ofm.wa.gov >

Hi Peter:

You'll get a mix of answers I'm sure. FWIW, I've got multiple SAS datasets online (PC's) that average around 3.5 million obs so records counts don't really matter.

The LRECL of these Tables (if converted to SAS) would, I think be more relevant.

Also, if the data can be edited/modified, then by converting the data to SAS before analysis would afford you an audit trail (i.e., even though the source data can change, you've always got a snapshot, albeit in SAS, of the data at a given point in time that you could drill back into.

Have fun-

================== W.W. Viergever Principal Viergever & Associates Sacramento, CA (916) 923-2355 ==================