[LON-CAPA-admin] sig figs & tolerance?

[Guy Albertelli II]

[ Charset ISO-8859-1 unsupported, converting... ]
>     Thanks much for the help on my other questions this week.  As a 
> token of my esteem, here's another question.  -grin-
> 
>     We have a problem that does the following:
> 
>     Pick random number between 2.1 and 5.9 in .1 increments.
>     Multiply result by 10^9.
>     Divide result by 4.7*10^-23.  Save as answer
> 
>     The settings, held over from this problem's original CAPA 
> incarnation, were 2% tolerance, 2 significant figures.
> 
>     One line from a particular student submission report:
> 
> ascheire 	gwu 	147820964 	none 	1.0e38 	1.0E+38 	INCORRECT 	1126752146 	1 	
> 
> 
>     Any thoughts?  One of our profs recalls that CAPA used to have a 
> similar problem where if the tolerance range did not cover the rounding 
> for sig figs, one could not possibly enter a correct answer.  We then 
> had to either increase sig figs or change the tolerance.
> 
>     Any thoughts?  In this case the random number was 4.8 so the 
> unrounded result is 1.021e36 so 1.0e36 is certainly outside the 2% 
> tolerance.
> 
>     I'm inclined to believe that we've run into this same problem.  Is 
> this expected behavior?  Should our perl script do the rounding for us?  
> Should I put this in as a bug?


The bug is with the problem. You have specified a sitiaution in which I
can't understand what you want. And thus I can't tell if the student
is right or wrong. (It may be obviou to you that the tol range isn't
as important ans the sig figs but how is lon-capa to know?)


Lon-CAPA 2.0 now sends a error message to the instructor when this
situation occurs.

(the "Computer's answer is incorrect ("..."). It is likely that the
tolerance range ... or significant figures ... need to be adjusted."
message.)


-- 
guy at albertelli.com   0-7-2-1-27,137