Date:         Tue, 15 Jan 2002 07:36:35 -0000
Reply-To:     Roland <roland.rashleigh-berry@VIRGIN.NET>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Roland <roland.rashleigh-berry@VIRGIN.NET>
Organization: Virgin Net Usenet Service
Subject:      Re: Validation of the SAS system, clinical setting

I wonder how stable SAS software is. I remember when we went to version 6 at
a place I work I was writing a chargeback system. I got a little suspicious
that the totals were not correct one month so I did a check. It turned out
that proc append was failing to append one in exactly 32 times without
giving a warning message and instead making out everything had worked. I
managed to distill the problem to its simplest form so that it was clear it
was failing to do the append exactly one in thirty two times and sent the
problem off to SAS. But they could not get it to repeat.

I would say, therefore, that the only good way to validate SAS output is to
duplicate program it in another language, preferably on a different
platform, and thoroughly compare the outputs. As for being sure that the SAS
system will never let you down - then don't ever assume this. And there is
no amount of testing you could ever do to completely validate it.

It's not that SAS is bad. It is a complex software package and things are
bound to go wrong on occassion. If you wanted more stable software then you
would have to use Cobol or RPG III, maybe.

"Don Hurst" <hurst_don@ALLERGAN.COM> wrote in message
news:200201142139.g0ELdQC208996@listserv.cc.uga.edu...
> I would be interested in any comments that SASser's might have regarding
> validation of the SAS system and hot fixes.  There is increasing pressure
> in the clinical study environment to perform IQ, OQ, and PQ of the SAS
> system within I.S. and also as it pertains to hot fixes.
>
> Any insight, discussion, or experiences related to the SAS system
> validation issue would be appreciated.
>
> Thanks,
>
> Don Hurst
> Manager, Statistical Systems
> Allergan, Inc