What's With The Brutal Curves At The Undergrad Level ?

<p>i thought i understood the curving process but now i'm just a bit confused.</p>

<p>so can someone give me a situation where a curve on a test would actually hurt those that received 100 (out of a 100) but benefited those who had lower grades?</p>

<p>and can someone give me a second situation where a curve on a test would actually hurt those that got the lower grades but benefit those who received the high grades, the 100s in the test, in order to differentiate their understanding of the material?</p>

<p>it really sounds like curving doesn't help those who score high on tests at all, just those that score low because it boosts their grade.</p>

<p>The students who perform significantly above the curve might end up with A+'s in the class - something that's not built into the curve, but rather, at the professor's discretion. The curve is not designed to hurt these people, and thus, it usually won't.</p>

<p>Here's how my vector calc professor explained the use of curving and giving exams with low means:</p>

<p>"If one is giving an exam with 100 possible points, why not use the entire range? If the lowest student gets a 90/100, that means that out of 100 points, the students only had to work for 10. On the other hand, if the low was a 20/100, the student has to work for 80 of the points. Curving allows me to give 10 problems of increasing difficulty and get an accurate picture of the class. </p>

<p>--one or two problems that anyone should get (to separate out the f's), a couple problems that you need a rudimentary handle on the course to get (for the d's and c-'s).</p>

<p>--about 5 problems that when added with the other 3, will take up the entire exam period for most students (the c-b range would be able to solve).</p>

<p>--two brutal questions that a b+ student might make some good assumptions on, an a- student will set up, an a student might get halfway through solving, and an a+ student will solve perfectly."</p>

<p>Giving an exam with a high mean, even when you have the ability to grade based on the performance allows you to use the entire range. In his 40 years of teaching, he had the art of exam making down to an exact science - when he handed out the exams, he predicted an average of 61. The real average? 60.5.</p>

<p>
[quote]
i thought i understood the curving process but now i'm just a bit confused.

[/quote]
</p>

<p>It's ok...it's still confusing to me sometimes...take a course in statistics and it'll make a lot more sense. </p>

<p>
[quote]
so can someone give me a situation where a curve on a test would actually hurt those that received 100 (out of a 100) but benefited those who had lower grades?

[/quote]
</p>

<p>
[quote]
I'm curious to how thoes 15 people that scored 100% on their tests would get different grades. assuming a 100% is the max you can get on the test, how would your 100% become a different score using a curve? a 100% means you got everything, so even in a curve wouldn't you still come out on top?

[/quote]
</p>

<p>No...if you scored a perfect grade on an exam, curving will never hurt. You're still at the top of the curve. Professors will never force the curve to take a perfect bell-shape so that people who aced the exam will get anything less than the top grade. But remember, most classes are usually not graded on one single exam....</p>

<p>
[quote]
it really sounds like curving doesn't help those who score high on tests at all, just those that score low because it boosts their grade.

[/quote]
</p>

<p>"High" is relative. If you scored 78/100 and this puts you 3rd in a class of 50, you scored high. And with a curve you'll get a grade that reflects how well you did compared to the class as a whole, instead of it being something that's decided by an arbitrary cutoff point that may or may not be suitable for that test (like, say, 80 = B- as is common in high school - why should one of the highest grades in the class get a C?).</p>

<p>If there are 15 people getting 100% on a college test in a reasonably normal-sized class, either the test was too easy or the class' abilities are really strangely distributed.</p>

<p>BigDaddy asked about AP scoring- I can tell you what I know.</p>

<p>There are some questions on the AP multiple choice parts that are reused for several years (this is why AP is so big about not talking about the test). They compare scores from year to year on the test to determine if a certain group of students is better or worse than average. The whole exam ends up being worth something in the 150-200 point range (I think, this is what I remember from my US history teacher 3 years ago), and the cutoffs for each number-score are adjusted to based on the data about how this year's students compare to the students of previous years, in order to keep a roughly constant level of acheivement needed for whatever score. On some tests, however, there is a a creep of cutoffs that make it more difficult to get high scores; I believe that the standards for the statistics AP test have changed significantly since it was created, partially because more can be expected since the teachers have taught it more and should be better now.</p>

<p>On free response sections, the most experienced graders read a bunch of responses. For math/science tests, they identify what the most important parts of each question are and assign a point value to each. A multi-part physics problem might be worth 8 points, where 1 point is stating F=ma and 1/2 point is plugging in the right numbers etc. Essays for APUSH and the English exams are graded on a 9 point scale. I think each gets read by two graders, and if they agree a score is asssigned; if not, it gets sent to a third person. For all subject areas, examples of papers of each score are selected and copied. All the other graders read them and refer to them when grading all the other essays.</p>

<p>The point value of each part of the test (MC/free response) is fixed by test, I think.</p>

<p>In short, AP grading is really crazy.</p>

<p>
[quote]
it really sounds like curving doesn't help those who score high on tests at all, just those that score low because it boosts their grade.

[/quote]
</p>

<p>These are both situations from math courses when I was in college.</p>

<p>I got a 97 on an exam where the average was a 92 or 93.</p>

<p>On another exam the next semester (new prof) I got a 98 on an exam where the average was in the low 40s.</p>

<p>On which exam did I clearly demonstrate I knew the material better than my classmates? Or if you don't like the comparison to classmates ... on which exam did I clearly domonstrate I knew the material compared to a very high absolute standard? Lower means allows the students who really know the material stand out. I don't buy the argument that on the test where the mean was 92% that the whole class knew the material dead on ... in my opinion the whole class knew the material pretty well and the test was pretty easy.</p>

<p>Ironically, having a curve also helps those on the lower half of the scale. If the mean is a 92 and you get a 20 or 30 then you are screwed. If the mean is a 50 and you get a 30 you still have a chance to recover through the term.</p>

<p>I'm pretty much convinced that tests with a mean of 50% and graded on a curve provide the best insight into the relative learning of all the students in a class.</p>