Date: Tue, 1 Oct 2002 15:43:19 +0100
Reply-To: John Whittington <John.W@MEDISCIENCE.CO.UK>
Sender: "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From: John Whittington <John.W@MEDISCIENCE.CO.UK>
Subject: Re: Last record in a dataset
In-Reply-To: <E17wIiv-0003ht-00@coumxnn01.netbenefit.co.uk>
Content-Type: text/plain; charset="us-ascii"; format=flowed
At 10:47 01/10/02 +0100, Peter Crawford wrote (in part):
>As put by "John Whittington <John.W@MEDISCIENCE.CO.UK>"
>"I suppose one could write code (with appropriate
>"error-trapping) which deliberately attempted to
>"read observations _beyond_ the end of the dataset
>"and then work backwards/fowards as necessary, using
>"point=, in order to identify the true 'last observation'
>
>Except that we don't need to look _beyond_ just backwards
A couple of points here.
Firstly, are you sure that there is no situation in which NOBS could
UNDER-report the number of observations? If such a situation could exist,
then there clearly would be a need to 'look beyond' point=NOBS to be sure
of finding the real 'last observation'.
Secondly, what you have been reporting leads me to wonder whether what I
was proposing would actually always work. What if the 'true last
observation' were one in which all of the variables had 'missing'
values? From what you're saying, I suspect it would not be possible to
identify whether it was 'a real last observation with missing values' or
not! I was expecting some sort of trappable error, not 'an observation
full of missing values' if one set POINT= to an 'invalid' number.
Kind Regards,
John
----------------------------------------------------------------
Dr John Whittington, Voice: +44 (0) 1296 730225
Mediscience Services Fax: +44 (0) 1296 738893
Twyford Manor, Twyford, E-mail: John.W@mediscience.co.uk
Buckingham MK18 4EL, UK mediscience@compuserve.com
----------------------------------------------------------------
| 
