home *** CD-ROM | disk | FTP | other *** search
- Comments: Gated by NETNEWS@AUVM.AMERICAN.EDU
- Path: sparky!uunet!wupost!zaphod.mps.ohio-state.edu!darwin.sura.net!paladin.american.edu!auvm!MAYO.EDU!WOLLAN
- Message-ID: <9301061416.AA13237@hsr>
- Newsgroups: bit.listserv.stat-l
- Date: Wed, 6 Jan 1993 08:16:44 CST
- Sender: STATISTICAL CONSULTING <STAT-L@MCGILL1.BITNET>
- From: Peter Wollan <wollan@MAYO.EDU>
- Subject: Re: Textbook formulas and computers
- Lines: 19
-
- John Wendell asks if Maple's evalf of an exact expression is an improvement
- over the evil textbook computational formulas--
-
- Probably not; it depends on how Maple goes about doing its evalf, and
- judging by the way numerical procedures have been tacked on to an otherwise
- wonderful package, it probably just sequentially evaluates the terms and
- combines them in order, which is disastrous. For example, evaluating a
- polynomial, the highest-order term is likely to kill everything else. Aside
- from that, which you probably can't do anything about, using evalf(",16)
- ought to be a little bit safer, and exactly the same speed, since it should
- be using 8-bite floating point arithmetic anyway. But again, I don't know
- what Maple will do there.
- The best numerical algorithm is going to be specific to the task (of course!).
- Weisberg's regression book has a good discussion of the sweep algorithm for
- regression (inversion of a positive semi-definite matrix)--it's a good place to
- start, to get an idea of what problems need to be addressed. Chapter 1,
- I think.
-
- Peter Wollan
-