RE: what good is "percentage error"?

[gastineau@badgerden.com (John E. Gastineau)]

On 12 May 97 at 11:09, JACK L. URETSKY (C) 1996; HEP wrote:

>  The guiding principle should be, I think, don't teach anything
>  until
> there is a payoff from learning it. 

Yes. Exactly why I'm asking the question. Why is anyone teaching 
"percentage error"? (aside from the chemists who grade that way...)

 Courses involving high
> precision instrumentation (usually more advanced courses) may
> involve multiple measurements of a quantity, in which case the
> results will differ from measurement to measurement.  In that case
> the student must resolve the issue: which measurement do I accept? 
> In that case, a major part of the course can be devoted to the study
> of statistics and statistical concepts, which can be used to help
> answer the students' question.

Does it really need to be an advanced course? It seems that this 
concept can be driven home by simple but multiple measurements of 
almost anything. Even before you tackle standard deviation, or 
especially if you don't, why not measure something ten times and look 
at the range of data? What's the real value? Is there a best value? 
All good questions. 

All this comes up as I am writing some exercises for CBL. Today's 
task is a photogate experiment for measuring "g" (and I won't start 
that old thread again by daring to name it!). Now, photogate 
measurements are not my favorites by a lot--but some people like them 
so I'm writing. Anyway, since the regression routines of calculators 
tend not to give you an uncertainty of the fitted quantities, why not 
just do the silly experiment ten times and look at that set of 
numbers? Students will learn a great deal more about 
experimentation, even at the HS level, and those last nine iterations 
of the experiment are totally painless given technology.

By the way, I think that experiment repetition is a long-ignored but
powerful benefit of doing experiments with computer or calculator
data acquisition. Repetition makes the fact that you are making an 
inexact measurement very very clear, and forces one to think about 
how to reduce that collection of results into one answer. 

> There was an earlier thread on this topic.

And I probably started it.

JEG

==================
John E. Gastineau                (304) 296-1966
900 B Ridgeway Ave.              gastineau@badgerden.com
Morgantown, WV 26505
www.badgerden.com/~gastineau