Date:         Fri, 12 Sep 2003 15:26:47 -0700
Reply-To:     cassell.david@EPAMAIL.EPA.GOV
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         "David L. Cassell" <cassell.david@EPAMAIL.EPA.GOV>
Subject:      Re: Excel inaccuracies
Comments: To: Steven.H.Feder@FRB.GOV
Content-type: text/plain; charset=US-ASCII

Steven.H.Feder@FRB.GOV wrote:
> I may be called on to run some stat procs to compare to
> Excel output, mayb a REG or two.
> Apparently Excel has the capability to run some stats
> within a spreadsheet. But I have heard that Excel may
> give inaccurate results in some cases.
>  . . . .

It appears that you have already done your homework, and that
you have already received some excellent advice from some of
the SAS-L gurus.  I would just like to point out that the
'multicollinearity' issue is likely to be your biggest headache
when running regressions.  Excel (and every other spreadsheet I
know, just to be fair) does a lousy job of inverting a matrix.
So any time your data matrix is ill-conditioned, you're likely
to run into problems.  And, since Excel does a lousy job of the
details on this, you may not be able to tell from Excel's results
that the matrix *is* ill-conditioned.. so you cannot tell when
you're likely to have trouble.  Just compare PROC REG against
Excel when you have built in substantial multicollinearity, and
you'll see.

Granted, there are plenty of other problems with stat algorithms in
spreadsheets...
David
--
David Cassell, CSC
Cassell.David@epa.gov
Senior computing specialist
mathematical statistician