home *** CD-ROM | disk | FTP | other *** search
- Comments: Gated by NETNEWS@AUVM.AMERICAN.EDU
- Path: sparky!uunet!spool.mu.edu!uwm.edu!psuvax1!psuvm!auvm!CCDOSH1.EM.CDC.GOV!DXA6
- Encoding: 48 TEXT
- X-Mailer: Microsoft Mail V3.0
- Message-ID: <2B4906FD@router.em.cdc.gov>
- Newsgroups: bit.listserv.stat-l
- Date: Mon, 4 Jan 1993 14:13:00 EST
- Sender: STATISTICAL CONSULTING <STAT-L@MCGILL1.BITNET>
- From: dxa6@CCDOSH1.EM.CDC.GOV
- Subject: Re: Spreadsheet Software
- Lines: 47
-
- A couple of weeks ago there was a discussion on the list about the perils of
- using spreadsheets for statistics. I didn't save the postings, but I do
- recall that at one point someone (Phillip Miller, I believe) responded to
- a query from a fellow CDC employee, Deanna Wild, with a message that said,
- "the formulas in the textbooks are wrong," or something to that effect, and
- that was why one should stick to software specifically designed to do
- statistics.
-
- At the time, I was too busy to think about it, but over the holidays I began
- to wonder, what formulas exactly are "wrong?" When I sat in Dr. Fleiss'
- class, I can't recall him ever saying that certain formulas in his book were
- in error, and that the correct ones could only be found in his notes. Ditto
- for Dr. Kleinbaum's classes. In fact, I can't recall any of my statistics
- professors over the years saying that the formulas in the books were wrong.
-
- Now, I admit that I'm not a statistician. Neither am I a conspiracy buff, or
- I might think that the "true and correct" formulas were hidden away in a safe
- somewhere, and that only the inner circles of the statisticians guild had the
- combination, while the rest of us were left wandering the dark. But I think
- that any claim that the textbooks are "wrong" ought to be substantiated.
- Textbooks do contain some computational formulas which generate
- approximations, but those are indicated as such. Besides, I was taught to
- use many of those "approximate" formulas exclusively when I was in
- engineering school, less than 20 years ago, when a pocket calculator that did
- logs and exponents was considered advanced.
-
- I confess I have used spreadsheet software to automate simple statistics that
- I used to calculate by hand: mean, median, standard deviation, standard error,
- normal-approximated confidence limits on a proportion, simple linear
- regression, and test statistics for z-tests, t-tests and one-way ANOVAs, to
- three or four significant digits without any difficulty. I used either the
- built-in formulas or set up the computations using formulas from textbooks.
- I don't use spreadsheets to calculate exact p-values and I wouldn't attempt
- to do or teach complex analyses using a spreadsheet.
-
- All tools and methodologies have their uses and limitations. It would be nice
- if we could all have a pocket computer that had built in SAS, SPSS, BMDP,
- SYSTAT, etc. and cost $99.95 with documentation, so that we could always do
- the best possible statistical computations for every problem. Flame me if
- you like, but if a textbook and a pocket calculator were sufficient under
- some circumstances, then a textbook and a spreadsheet program can be
- sufficient sometimes as well.
-
- David R. Arday, B.S.Ch.E., M.D., M.P.H.
- Office on Smoking and Health
- Centers for Disease Control
- dxa6@ccdosh1.cdc.em.gov
-