mwm@ea.UUCP (09/24/84)
Moved from net.flame(.tepid):
/***** ea:net.flame / cbnap!adm / 5:36 pm Sep 21, 1984 */
Taking a freshman physics course at Purdue, I recieved 35% on the first test.
When the curve for this exam was posted, this turned out to be a "C". This
was typical of many tests for that course not only for our group of freshman,
but it had been true for years before I got there, and not only for that
course, but others as well (although somewhat less dramatic). I've also talked
with other students from other schools and this, by no means, is limited to
Purdue but is an affliction that hits many school (most of them I hear about
are big name schools, but that could be just because they are more visible).
In my mind, their is no excuse for this. No matter what can be said about the
students not working hard enough (etc.) with scores this dramatic some blame
*must* be put on the administration. Is it that the kids aren't learning the
material, or are the tests bad? Hard to tell I guess, but in a situation such
as that, I would feel the same way about cheating as those interviewed did!
Sometimes I'm just clearing my throat...
Craig Votava
AT&T Bell Laboratories, Columbus
...ihnp4!cbnap!cmv
/* ---------- */
That brings up a whole different kettle of fish: The grading scale used
by schools. CalTech was similar to Purdue. Here at Oklahoma University,
it's closer to the "cannonical" curve:
A 90-100
B 80-90
C 70-80
D 65-70
F <70
This seems somewhat silly. If 90% of your students do reasonably well
(C or better), you've just crammed 90% of the data into 30% of the scale.
You've got lots of accuracy for people with F's, and very little for the
best. Something like Purdue's scale would make more sense (lots of accuracy
for the very best) for a "good" school. Even better, something like:
A 80-100
B 60-80
C 40-60
D 20-40
F 0-20
Would seem to give you the same accuracy for all types of students.
Comments, anyone?
<mikejab@uokvax.UUCP (09/27/84)
/***** uokvax:net.college / ea!mwm / 3:32 pm Sep 24, 1984 */ A 80-100 B 60-80 C 40-60 D 20-40 F 0-20 /* ---------- */ This is a really kinda strange grading scale, also. Are you trying to tell me that a student who misses 4 questions out of 5 still passes the class? That doesn't sound like a reasonable system, either. There is something to be said about "mandatory bell curve" methods of grading: I don't agree with that scheme, but it helps with this particular problem. To be realistic, you would really like to grade the questions on the test: i) This question tests knowledge of the material (able to recite theorems, apply directly, able to work test problem very similar to homework problem) ii) This question tests understanding of the material (able to apply theorems in non-obvious ways, able to generalize or mimic theorems, able to work test problem by applying homework problem) Then make the grading scale relative to the questions: i) To receive a passing grade, you must pass 80% of the 'knowledge of the material'; to receive an 'A' or 'B', you must pass 90% of the 'knowledge of the material' questions. ii) For the second kind of question, you must pass 25% - D 50% - C 75% - B 85-90% - A That way, a "passing grade" means that you know MOST of the material in the course, at least on a casual basis; a high grade means that you understand most of the material in the course. Of course, I wouldn't like to be the one grading the questions. Jeff Bowles Lisle, IL
mwm@ea.UUCP (09/27/84)
/***** ea:net.college / uokvax!jab / 8:58 pm Sep 26, 1984 */ /***** uokvax:net.college / ea!mwm / 3:32 pm Sep 24, 1984 */ A 80-100 B 60-80 C 40-60 D 20-40 F 0-20 /* ---------- */ This is a really kinda strange grading scale, also. Are you trying to tell me that a student who misses 4 questions out of 5 still passes the class? That doesn't sound like a reasonable system, either. Jeff Bowles Lisle, IL /* ---------- */ Depends on how hard the test is. If it is sufficiently difficult, getting 2 or three right could indicate that you knew the material, but weren't proficient at using it (a C grade), and one right would mean that you were had some grasp of the material - and possibly deserved to pass. The grade scale must of course reflect the difficulty of the test. Maybe I should have said that the current crop of tests don't differentiate among the better students, but only among the worst. Of course, I can't write tests that fit some magic curve, so I don't worry about what score gets what grade until I've got all the grades in hand. <mike
scot@dartvax.UUCP (Scot Drysdale) (10/01/84)
estions
require applying principles to new situations. Some of these applications are
easy ones, most are moderately difficult, and usually one is taken from a
recent research paper so is very hard. On this type of exam 80% can be a
solid A and 50% a respectible C. (The handful of students who get close to
100 get rave recommendations to graduate schools.) The choice of questions
and partial-credit policies can make the numerical scale come out wherever
the professor chooses. The measure of the quality of an exam is how well
it measures student understanding, not whether the outcome follows some
pre-ordained distribution. (I do think that students should have some idea
of how difficult you think the exam is and what your expectations are before
they take it, so that they do not panic when they cannot solve half the
questions on an exam where you expect most students to be unable to solve
half the questions.)
Scot Drysdale (scot @ dartmouth)