home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!pipex!bnr.co.uk!uknet!mcsun!Germany.EU.net!urmel.informatik.rwth-aachen.de!rama!dak
- From: dak@rama.informatik.rwth-aachen.de (David Kastrup)
- Newsgroups: sci.math
- Subject: Re: Rounding Rules
- Message-ID: <dak.726242739@rama>
- Date: 5 Jan 93 14:05:39 GMT
- References: <1992Dec20.003018.14325@sol.ctr.columbia.edu> <1992Dec23.140148.21009@hubcap.clemson.edu> <1993Jan4.142226.9639@iscsvax.uni.edu>
- Organization: Rechnerbetrieb Informatik - RWTH Aachen
- Lines: 31
- NNTP-Posting-Host: rama.informatik.rwth-aachen.de
-
- kraai4712@iscsvax.uni.edu writes:
-
- >In article <1992Dec23.140148.21009@hubcap.clemson.edu>, steve@hubcap.clemson.edu ("Steve" Stevenson) writes:
- >> In article <1992Dec20.003018.14325@sol.ctr.columbia.edu> shaw@toadflax.UCDavis.EDU (Rob Shaw) writes:
- >>>What is the rationale behind the following rounding rule?
- >>>
- >>>when dealing with 5's followed by all zero's, check the
- >>>next digit to the left. If it's even, round down; odd,
- >>>round up.
- >>>
- >>>For example both 1.13500 and 1.14500 are 1.14 to 3 places.
-
- >My personal favorite rounding rule is to round anything that has {0,1,2,3,4} in
- >the "rounding place" "down" and anything that has {5,6,7,8,9} "up". So as in
- >the previous example, 1.13500 would be rounded to 1.14 and 1.14500 would be
- >rounded to 1.15.
-
- >To this my teachers always gasped that this method was biased, and no amount of
- >pursuasion using the above-mentioned sets could pursuade them.
-
- >I have come to believe that the people who thought this was biased believed
- >that anything that had a zero digit in the place to be rounded couldn't be
- >rounded or didn't need to be rounded or some other mental block.
-
- One problem with your rounding method is that rounding errors might add up.
- So, successive rounding of 1.44445 yields 1.4445, 1.445, 1.45, 1.5, 2.0.
- Rounding to nearest even will, when doing successive rounding, never
- have an error more than half of the then LSB. Also, there are SOME
- more properties of the real group that hold valid, making numerical
- computations slightly more reliable. See Knuth, Seminumerical Algorithms
- (Art of comp. progr., Vol 2) for reasons and rationalizations with proofs.
-