Date:         Fri, 22 Jun 2007 18:06:01 -0400
Reply-To:     Joe Whitehurst <joewhitehurst@GMAIL.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Joe Whitehurst <joewhitehurst@GMAIL.COM>
Subject:      Re: DATETIME conversion by SAS 9
Comments: To: "Audimar P. Bangi" <audi@sas2themax.com>
In-Reply-To:  <622B06615F3243898AC2E5F7CB36BF9F@WASILLA>
Content-Type: text/plain; charset=ISO-8859-1; format=flowed

Audi,

Check out the Physics Guy Rap:
http://blog.sciam.com/index.php?title=rapping_about_physics&more=1&c=1&tb=1&pb=1.
Now what I would really like to hear is a SAS Gal/Guy Rap!  Or a
Shakespearean Rap.  Could you do it?  I'll give you the first line: " SAS to
the max no matter whom I have to axe."

Joe


On 6/22/07, Audimar P. Bangi <audi@sas2themax.com> wrote:
>
> Alan,
>
> >Very few people, IMO, would change a date value to a datetime.
>
> Yes, very few, if any, would do that, but that's a possibility. For
> example,
> if you are joining or merging datasets with mixed DATE/DATETIME formats,
> you
> would probably want to do a conversion to a common date format first. The
> conversion offered by the SAS editor would mess things up big time.
>
> >It means changing all of the values in the data being analyzed.
>
> Not really -- only the column/variable that is currently being edited.
>
> >Why? If it is a date value then why would you want every value to reflect
> >midnight? Similarly, why change a datetime to a date? You would lose
> >granularity.
>
> If SAS doesn't round off days to whole numbers, a double-precision numeric
> value should be able to store the date part and the time part for a
> DATE-formatted variable. You can then change from DATE to DATETIME (and
> vice
> versa) without losing any significant fraction of the time value. But SAS,
> by design, doesn't do any internal conversion for this type of format
> transformation inside its Data Editor, which, to me, is not a good
> architectural decision.
>
> >...you leave the values alone and just change how they are presented.
>
> As I mentioned in my previous post, this behavior yields unrealistic and
> weird-looking results.
>
> > To do a proper conversion requires a pass of all of the data and a
> change
> > in the underlying values.
>
> Only the column/variable that is being edited needs to be changed (when
> the
> user clicks on the "Apply" button). Because the new format after the
> change
> has been applied is persisted with the dataset, you get correct values all
> the time.
>
> Regards, Audi
>