home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: rec.boats
- Path: sparky!uunet!usc!sdd.hp.com!ux1.cso.uiuc.edu!news.cso.uiuc.edu!vmd.cso.uiuc.edu!EPLUS17
- From: EPLUS17@vmd.cso.uiuc.edu (Richard Engelbrecht-Wiggans)
- Subject: Re: Celestial vs. GPS (was Re: Bermuda Trip - 7/'93)
- References: <1992Dec15.213557.17851@atlastele.com> <1hqgrhINNlcn@calvin.NYU.EDU> <29093@oasys.dt.navy.mil>
- Message-ID: <168CDCF52.EPLUS17@vmd.cso.uiuc.edu>
- Sender: usenet@news.cso.uiuc.edu (Net Noise owner)
- Organization: C.C.S.O.
- Date: Wed, 30 Dec 1992 20:44:29 GMT
- Lines: 58
-
- In article <29093@oasys.dt.navy.mil>
- hanners@oasys.dt.navy.mil (Robert Hanners) writes:
-
- >In rec.boats, roy@mchip00.med.nyu.edu (Roy Smith) writes:
- >>bartley@spss.com (Dennis Bartley) writes:
- >>
- >> I think part of the problem is that most people don't understand the
- >>difference between precision and accuracy, and assume that just because a
- >>Loran shows you your position to 1/10th (or even 1/100th) of a minute it
- >>must be correct.
- >
- >
- >My question is this: Why isn't it reasonable for the unit to calculate and
- >display a measure of this potential error, as individual values (for
- >the discerning user :-) ) or as a total? Or is this in fact a feature
- >of some makes? The error itself could be a statistical measure such
- >as 50%, standard deviation, 2 x sd, or whatever. Preferrably such
- >weighting would be user selectable.
- >
- At best, the unit could display an estimate of KNOWN error sources.
- All too often in my experience, we overlook important sources of
- error. So, having the unit display an estimate of error would tend
- to further confuse precision with accuracy.
-
- An example: A friend--engineer by profession--reported that his new
- car got 19.328 miles per gallon, plus or minus 0.002. How did he
- calculate this? He filled up his tank three times, keeping track
- of the distance he had driven and divided that by the amount of
- gas purchased.
-
- But he only knew his total miles (about 1000) to plus or minus
- about 0.1 miles; an error of about 1 in 10,000. The fuel pump
- reports the amount of gas delivered to the nearest 0.001 gallons;
- for a 16 gallon tankful, this gives an error of about 1 in 16,000
- for each tankful. Since some of the errors might offset each other,
- he figured a total "typical" error of about 1 in 10,000, or 0.002
- out of about 20.
-
- Then I come along and ask "how do you know that your gas tank had
- exactly (to 0.001 gallons) the same amount in it at the beginning
- and end of the experiment?" Answer: he always uses the same pump
- and lets it shut itself off. Well, the next two fillups, I went
- along. After the pump shut itself off, we carefully topped the
- tank to the top of the fill tube, noting how much extra gas was
- delivered in the process (and siphoned some off before driving off).
- On the two fillups, the amount of gas needed to top of the tank after
- the pump had shut off differed by just over one gallon! In fact,
- he only knew that he had used about 45 plus or minus one gallons of
- fuel, and his mpg estimate was at best accurate to plus or minus
- one half mpg. He had overlooked an error source that completely
- swamped the errors that he had considered.
-
- How many boaters do you really think would remember that the error
- estimate is only an estimate and might be orders of magnitude off?
- I suspect that most would deceive themselves into thinking that
- they knew things more accurately than they really did.
-
- ..Richard E+17
-