Wow my class was dumb!

[Balain]

So I got my grade back from my final exam in my stats class. I got 85.4%. He didn't post the class average like the other exams He did post the total class average 37% Wow

[Starwulf]

Question is, does that frequently happen, or was it a one off? I remember when I was younger, I had a buddy that was in college, and the first day of one of his classes, his professor said "You will all fail 90% of the tests in here. I make them intentionally hard to promote more learning and critical thinking, even during test taking. Your actual grade will be in comparison to the rest of the class(So if class average was 36%, and you had 32%, You'd get an A). Some professors are weird like that I guess.

[thubby]

well, realize a good chunk of that is going to be 0's from no-shows and people who didnt drop.


Question is, does that frequently happen, or was it a one off? I remember when I was younger, I had a buddy that was in college, and the first day of one of his classes, his professor said "You will all fail 90% of the tests in here. I make them intentionally hard to promote more learning and critical thinking, even during test taking. Your actual grade will be in comparison to the rest of the class(So if class average was 36%, and you had 32%, You'd get an A). Some professors are weird like that I guess.

this is also completely unethical.

[WarKitty]

I've TA'd before. I really wish they let me do that with a class or two.

[Spiryt]

I think that inebriating them and selling them to coffee plantation in Brazil is the only solution now.

[Balain]

He did make all the exams hard. He made the final open book. I suspect many didn't study and ran out of time trying to look everything up.

People in class I talked after final thought it went okay for them. The averages for other exams were in the range of 60% to 70% So seems that many bombed the final to me.

[Eldan]

well, realize a good chunk of that is going to be 0's from no-shows and people who didnt drop.



this is also completely unethical.

How is making tests hard unethical?

[Douglas]

How is making tests hard unethical?
He might be referring to the part where your grade is determined by comparison to the rest of the class. Such a mechanism provides incentive to give no help to your classmates, or even to actively sabotage them, and can result in unfair grading standards due to random variation in the quality of your class.

I strongly agree that the best tests are ones where few or no people will ever actually get the theoretical maximum score, and full access to reference material is allowed but isn't enough to make up for lack of proper understanding. In my experience, these traits are strong indicators of the highest caliber of tests, test-writers, and teachers. Grading by comparison to the rest of the class, however, is an easy shortcut that gives perverse incentives and high variance in grading standards.

[Elder Tsofu]

We had a similar system in our schools a few decades ago, a parent schooled in that era vehemently complained that her daughter got bad grades since we had too many smartasses in the class. She shut up when it was explained to her that nowadays, theoretically speaking, everyone in class could get the top grade.

A good test is something as the above by douglas, it should somehow be representative to what you're supposed to know after the class...
Quite a few of our tests felt too easy, possibly because the University got more cash from the government if they had passing students rather than failing ones.

[Blue1005]

He might be referring to the part where your grade is determined by comparison to the rest of the class. Such a mechanism provides incentive to give no help to your classmates, or even to actively sabotage them, and can result in unfair grading standards due to random variation in the quality of your class.

I strongly agree that the best tests are ones where few or no people will ever actually get the theoretical maximum score, and full access to reference material is allowed but isn't enough to make up for lack of proper understanding. In my experience, these traits are strong indicators of the highest caliber of tests, test-writers, and teachers. Grading by comparison to the rest of the class, however, is an easy shortcut that gives perverse incentives and high variance in grading standards.


Correct me if I am wrong, but doesn't life compare you to others? To grade any way but is moronic. That is the way that 90% of the class gets 80% or better. The bell curve is there for a reason, too few actually use it.

[WarKitty]

He might be referring to the part where your grade is determined by comparison to the rest of the class. Such a mechanism provides incentive to give no help to your classmates, or even to actively sabotage them, and can result in unfair grading standards due to random variation in the quality of your class.

I strongly agree that the best tests are ones where few or no people will ever actually get the theoretical maximum score, and full access to reference material is allowed but isn't enough to make up for lack of proper understanding. In my experience, these traits are strong indicators of the highest caliber of tests, test-writers, and teachers. Grading by comparison to the rest of the class, however, is an easy shortcut that gives perverse incentives and high variance in grading standards.

It depends on what "comparison" means. Comparison doesn't necessarily mean that the grades are strictly curved. When I TA'd I often did use comparison to gauge if there were things that the class just wasn't getting (which made it far more likely that I'd simply communicated it badly than if a single student wasn't getting it). Outliers were typically ignored, though.


Correct me if I am wrong, but doesn't life compare you to others? To grade any way but is moronic. That is the way that 90% of the class gets 80% or better. The bell curve is there for a reason, too few actually use it.

Thing is you have to do it intelligently. For example, I had one summer class where all my students earned A's. And I felt at the end that they had legitimately earned them. That class was just better than other classes. It would be neither fair nor realistic to grade some of the one class lower because other students in the class were good.

The grading system is comparing you to others. It's just comparing you to some standard that's what the average student can do, instead of what specifically the ones around you can do.

[Blue1005]

It depends on what "comparison" means. Comparison doesn't necessarily mean that the grades are strictly curved. When I TA'd I often did use comparison to gauge if there were things that the class just wasn't getting (which made it far more likely that I'd simply communicated it badly than if a single student wasn't getting it). Outliers were typically ignored, though.



Thing is you have to do it intelligently. For example, I had one summer class where all my students earned A's. And I felt at the end that they had legitimately earned them. That class was just better than other classes. It would be neither fair nor realistic to grade some of the one class lower because other students in the class were good.

The grading system is comparing you to others. It's just comparing you to some standard that's what the average student can do, instead of what specifically the ones around you can do.


I too am an educator, and I must ask, which one is more realistic in regards to life? I understand your point, but for everyone to actually get an A means that there is no bell curve, what grades were initially based on.

[Ashtagon]

A single class is just a vanishingly small part of their cohort that is doing the same test across the country. It is quite possible for everyone in one class to legitimately deserve an "A" without affecting the overall bell curve.

[Emmerask]

Question is, does that frequently happen, or was it a one off? I remember when I was younger, I had a buddy that was in college, and the first day of one of his classes, his professor said "You will all fail 90% of the tests in here. I make them intentionally hard to promote more learning and critical thinking, even during test taking. Your actual grade will be in comparison to the rest of the class(So if class average was 36%, and you had 32%, You'd get an A). Some professors are weird like that I guess.

I would just urge everyone to write nothing at all in the exam, since everyone has 0%, 0% is an A :smallwink:


So I got my grade back from my final exam in my stats class. I got 85.4%. He didn't post the class average like the other exams He did post the total class average 37% Wow

stats in general is not a hard class really, it actually was the simplest math class I had at university and the avg was somewhere around the 70% mark, so with such a low average more then likely your prof is just plain and simple not good at his job.

[valadil]

I would just urge everyone to write nothing at all in the exam, since everyone has 0%, 0% is an A :smallwink:

Sounds a lot like the prisoner's dilemma. Would you really trust all those classmates not to get greedy and screw you over? Sure everyone benefits if the whole class gets a 0, but someone might feel it's in their best interest to get an A and leave everyone else behind.

As a student I always liked the strategy of making tests too hard to ace. But I wasn't the type of student to try and game the system.

[WarKitty]

I too am an educator, and I must ask, which one is more realistic in regards to life? I understand your point, but for everyone to actually get an A means that there is no bell curve, what grades were initially based on.


A single class is just a vanishingly small part of their cohort that is doing the same test across the country. It is quite possible for everyone in one class to legitimately deserve an "A" without affecting the overall bell curve.

Keep in mind TA classes are fairly small. And summer students I've found are almost universally better than year students. So we're talking a class of under 10 students. My largest are under 30. The probability of a class of that size actually fitting a bell curve is vanishingly small. Trying to impose a bell curve on it would amount to simply introducing a significant random element into grading.

[Teddy]

We had a similar system in our schools a few decades ago, a parent schooled in that era vehemently complained that her daughter got bad grades since we had too many smartasses in the class. She shut up when it was explained to her that nowadays, theoretically speaking, everyone in class could get the top grade.

Yeah, mum is still a bit bitter about the class where she had her grade lowered just because the teacher ran out of top grades...


A good test is something as the above by douglas, it should somehow be representative to what you're supposed to know after the class...
Quite a few of our tests felt too easy, possibly because the University got more cash from the government if they had passing students rather than failing ones.

I would agree, but I guess I'm a bit blind toward difficulty, because the amount of coursemates I have who yet fail them is a bit staggering, even when they actually study...


I too am an educator, and I must ask, which one is more realistic in regards to life? I understand your point, but for everyone to actually get an A means that there is no bell curve, what grades were initially based on.

To build upon what Ashtagon and WarKitty said, when sampling statistics, you're generally required to have a sample size of 1000 subjects selected from different backgrounds (minus a few who couldn't respond) in order to minimise the margin of error. For a normal class, it isn't outside the realm of possibility for 50% of the students to be unusually intelligent out of pure happenstance (and it may even be likely with some prestige schools). The same thing would never happen on a national scale.

And consider two classes, where one has a great teacher and the other one a terrible one. In this case, the people with the best grades in the second class may only be on par with the ones with okay grades in the first one in terms of acquired knowledge.

[MikelaC1]

It was a stats class. Everyone is supposed to fail at stats the first time out. Then they first hand knowledge of how the bell curve works.

[Jay R]

It was a stats class. Everyone is supposed to fail at stats the first time out. Then they first hand knowledge of how the bell curve works.

If they all failed, then that isn't how the bell curve works.

If you and person X both show the same level of ability and knowledge, then you should both the the same grade, even if your class had a bunch of geniuses and person X's class had a bunch of about-to-be dropouts.

You determine your grade; not the people around you.

[Drakeburn]

Relating to the thread title here, I remember around my first year in High School when my Freshman math class were given a Sudoku to work on. And the funny part was that I was probably the only one in my math class who can do a Sudoku, yet alone know what a Sudoku is. :smallbiggrin:

You see, before I entered High School, my mother used to be a big fan of Sudoku puzzles, so she taught me how to do them. So that is how I became familiar with the word.

[Yora]

Question is, does that frequently happen, or was it a one off? I remember when I was younger, I had a buddy that was in college, and the first day of one of his classes, his professor said "You will all fail 90% of the tests in here. I make them intentionally hard to promote more learning and critical thinking, even during test taking.
Last semester, the teacher in our Japanese gramar class announced a test on a fairly simple gramatical element for the next week. That next week we did the test and immediately went to grade each others test based on the correct answers he told us. That week 16 out of 21 got an F. The other two classes he teaches did just as bad.
The next week we did exactly the same test again just with different words, and that time there were still 10 people with an F. But four with an A, which meant at least one person went from an F to an A in just one week.
And of course, the week after, we did the test again with only two Fs and lots of As and Bs. That really was quite motivating to just sit down and learn that stuff, which really wasn't so hard, but will become terribly important in later semesters.

[Karoht]

Grading by comparison is rather pointless.
An education is about the individual. The grade should reflect how that individual performed, and should infer nothing about the rest of the class. Otherwise that grade is suspect. The grade should tell you how well I know that content covered by the course.

If I score 100/100 on a test and you score 50/100, my grade on that test should reflect that. If we must compete, we can compare grades after. Your grade will be lower. The comparison/competition is already there. Altering the grades based on someone elses performance is outright absurd.

I should not be able to get a passing grade with only 20/100 on a test, just because the rest of the class only scored an average of 36/100.

Then again, this also gets into testing and education/evaluation methods. Don't get me wrong, I place much of the responsibility on the shoulders of students and parent. But I scoff when we place paying customers in the hands of professors who maybe show up to teach class twice a semester and leave the TA's to do all the legwork.

[Deathslayer7]

Grading by comparison is rather pointless.
An education is about the individual. The grade should reflect how that individual performed, and should infer nothing about the rest of the class. Otherwise that grade is suspect. The grade should tell you how well I know that content covered by the course.

If I score 100/100 on a test and you score 50/100, my grade on that test should reflect that. If we must compete, we can compare grades after. Your grade will be lower. The comparison/competition is already there. Altering the grades based on someone elses performance is outright absurd.

I should not be able to get a passing grade with only 20/100 on a test, just because the rest of the class only scored an average of 36/100.

Then again, this also gets into testing and education/evaluation methods. Don't get me wrong, I place much of the responsibility on the shoulders of students and parent. But I scoff when we place paying customers in the hands of professors who maybe show up to teach class twice a semester and leave the TA's to do all the legwork.

I think this depends entirely on the course taken. English I would certainly agree. Engineering classes, I disagree.

I took Heat Transfer and the average on that test was about a 27%. The test scores went like this:

98, 92, 84, 69 (my score), 57 and then they dropped below that. Note this was probably a class size of about 50-60. Obviously 95% of the class failed that exam. It doesn't mean we didn't learn anything. Engineering classes are meant to be hard. Most professors understand this and compensate for it. If they were to go by the methodical grading scale of 93 or above is an A, 90 -93 A-, etc. Then a lot more people would fail engineering.

Sometimes it is just bad professors. I have had two classes where the professors taught horribly and graded harshly and failed over half the class. Still learned a lot though.

[The Succubus]

My class was pretty dumb too. I'll pick something other than Monk next time.

[Jade_Tarem]

I too am an educator, and I must ask, which one is more realistic in regards to life? I understand your point, but for everyone to actually get an A means that there is no bell curve, what grades were initially based on.

The "that's how it is in real life" justification is the product of lazy thinking. Competition exists in real life, therefore competition in the classroom is always a good thing, no matter what form it's presented in. This is false, as is the assertion that the bell curve is worth preserving just because that's how it was done in the past.

In any engineering, scientific, medical, or other constructive field (and many branches of service or entertainment industries), cooperation between a given employee and their coworkers is not only a good thing but absolutely required. Performance in many cases is determined by the performance of the group. If another engineer on my team is struggling with something, it is in *my* best interests to help him or her because, while letting them fail might make me look (slightly) better by comparison in the short term, everyone on the team suffers long-term if we fail to meet delivery deadlines and lose a contract. Yes, competition is a real thing, but you're competing *with* a team against *other* teams in most jobs. Yes, you compete for raises individually, but again, actual professionals understand that mentoring or aiding coworkers actually makes them look better to management, even though it theoretically narrows the skill gap between themselves and a more junior employee.

As an engineer and an educator myself, my observation is that the hallowed, venerable, obsolete, and vastly overrated bell curve system can only foster cooperation by forcing it on the students (say, with a group project), instead of encouraging it. Academic honesty is important, of course, but the mighty curve leaves no room for group study, sharing ideas, reasoned argument or debate, or collaborative research, all of which are more important than hammering into the skulls of kids that they should hate everyone smarter than they are because the very existence of such people is a threat. Even within the context of the oh-so-scary real world, the curve system is poison.

[valadil]

As an engineer and an educator myself, my observation is that the hallowed, venerable, obsolete, and vastly overrated bell curve system can only foster cooperation by forcing it on the students (say, with a group project), instead of encouraging it. Academic honesty is important, of course, but the mighty curve leaves no room for group study, sharing ideas, reasoned argument or debate, or collaborative research, all of which are more important than hammering into the skulls of kids that they should hate everyone smarter than they are because the very existence of such people is a threat. Even within the context of the oh-so-scary real world, the curve system is poison.

How about the sliding bell curve? Actually I don't know if it has a name, I just remember a few teachers describing it this way.

So you do the bell curve thing to figure out each student's relative position to the rest of the class. But the middle of the curve is variable and the teacher can assign it to whatever they feel the class earned. In an average class, the middle of the curve is a C. In a good class it's a B. Etc. Now it is in your best interest to help the rest of the class because it can bring up your grade as well.

[Jade_Tarem]

How about the sliding bell curve? Actually I don't know if it has a name, I just remember a few teachers describing it this way.

So you do the bell curve thing to figure out each student's relative position to the rest of the class. But the middle of the curve is variable and the teacher can assign it to whatever they feel the class earned. In an average class, the middle of the curve is a C. In a good class it's a B. Etc. Now it is in your best interest to help the rest of the class because it can bring up your grade as well.

While better than the standard curve, I fail to see how that's superior to the most objective measurement possible. Or no grades at all.

What you're describing is more common these days, largely among teachers or professors who either don't know how to write a good test or secretly like it when all of their students fail, ("Everyone failed the 30 page, 45 minute written exam! That means that the subject matter is super hard and I must be *really* smart if I'm the one teaching it!"), but know that the students, parents, and university will all get on their case incessantly if they hand out a big string of D's and F's. It's still not a better system than setting up a proper syllabus and grading scheme and giving out the grades the students actually earned.

[Balain]

All this reminded me of a conversation with a prof. Testing is the worst way to see if you know what you are doing or not. There is no way to really get away from tests though.

[thubby]

How is making tests hard unethical?


He might be referring to the part where your grade is determined by comparison to the rest of the class. Such a mechanism provides incentive to give no help to your classmates, or even to actively sabotage them, and can result in unfair grading standards due to random variation in the quality of your class.

^this


Correct me if I am wrong, but doesn't life compare you to others? To grade any way but is moronic. That is the way that 90% of the class gets 80% or better. The bell curve is there for a reason, too few actually use it.

school exists to teach, your grade reflects your mastery of the material. GPA is already ranked, so grading off of other students' performance is just compounding the comparison and actually removes any ability of outside observers to judge the quality of the class or school as a whole.

"more realistic" is always an ad-hoc excuse for bad habits. worse than that, the thing in question is never actually more realistic.
firstly in this case you're trying to ignore that the students are in fact customers. and secondly that in the real world cooperation is encouraged. helping your fellow employees is a sign of leadership that any company will reward.

[Reverent-One]

school exists to teach, your grade reflects your mastery of the material. GPA is already ranked, so grading off of other students' performance is actually compounding the comparison and actually removes any ability of outside observers to judge the quality of the class or school as a whole.

You can lose that ability without using curving as well through the differences in how difficult different classes are, or how well different teachers can put together tests that accurately measure learning the material covered. If the 90% of the class fails a test, even students that do well otherwise, it's probably either the professor's or test's fault, not the students'.

[Haruki-kun]

How is making tests hard unethical?


Correct me if I am wrong, but doesn't life compare you to others? To grade any way but is moronic. That is the way that 90% of the class gets 80% or better. The bell curve is there for a reason, too few actually use it.

Yeah, I disagree, sorry.

Suppose there are two sections of this class, Class A and Class B.

By mere randomness, Class A has 9 straight A students who usually score within the 95-100% range and one average student who can usually pull off 80-85%.

Class B has 10 pretty bad students who can barely get past 50%. Class B's average ends up being 35-40%. One or two students falls well below that average, but most of them walk out of there with an A to A- because they were all equally bad.

Meanwhile, in Class A, that one student walks out with a B- to C+ even though in reality he scored twice as much in the same class as the people in the other section. Life constantly compares you to others, but Class A's students will all show off that they got an A because "compared to everyone else in my class I'm really good." Whereas Class B's student will be hit pretty hard for not being as good as the Straight A students in his class.

Yes, I know, we're talking about a stats class, and statistically this scenario is unlikely to happen exactly as is (especially considering the teacher engineers the test to make sure people score low), but I think it helps illustrate why I don't like this system. Life will compare me to others on its own, and when it does I want to be able to show life that I was graded based on my own skills, not based on the skills of the people around me.

One of my peers told me a teacher talked to him about his work and told him off for slacking. He replied that he was still better than most of the people in the class. The teacher told him "If you want to make yourself feel better, go look at your classmates' work and compare it to yours. If you want to get a job, get back to work."

[Reverent-One]

Yeah, I disagree, sorry.

Suppose there are two sections of this class, Class A and Class B.


Counter-scenario:

Suppose there are two sections of this class, Class A and Class B, taught by different professors.

Professor A is fairly good at writing tests, though tends to lean on the easier side of the spectrum and to be a bit easier going in regards to partial credit. Professor B isn't as good at hitting the exact difficulty he's aiming for, or perhaps thinks that greater challenges prompt more actual learning, and agrees with you about curving.

Despite getting similar grades in other classes in the same field, most of class A gets scores in the 80-95% range, while only 1 or 2 get over 80% and only a handful end up with even a 75% in Class B. Those who do score higher in Class B probably know as much, if not more, than those getting 90-95% in Class A, but simply because of the difference in teachers, score much lower. That seem like a fair or accurate measurement to you?

I'd wager differences in teachers come up more than your hypothetical balances of all the high scoring students in one class and lower scoring ones in the other.

[Haruki-kun]

*snip*

Fair point. It does happen quite often.

I'm still not sure I'm convinced of it, though. In such a case I'd rather explore other options to make grading as even as possible. But then, I'm not an educator, so I may not be the best person to come up with it.

[Reverent-One]

Fair point. It does happen quite often.

I'm still not sure I'm convinced of it, though. In such a case I'd rather explore other options to make grading as even as possible. But then, I'm not an educator, so I may not be the best person to come up with it.

Another thing to consider is how easy to self-correct the two scenarios are. To use yours, assuming the sections are indentical (same teacher, material, ect) other than the students, the teacher could easily notice the significant discrepency on scores and use the scores of both sections to generate the curve. On the other hand, if the inequalities come from differences in teaching/test writing styles of different teachers...well, people can be very stubborn about whether their way is the right way or not.

[Haruki-kun]

Another thing to consider is how easy to self-correct the two scenarios are. To use yours, assuming the sections are indentical (same teacher, material, ect) other than the students, the teacher could easily notice the significant discrepency on scores and use the scores of both sections to generate the curve. On the other hand, if the inequalities come from differences in teaching/test writing styles of different teachers...well, people can be very stubborn about whether their way is the right way or not.

Actually, I considered that. It's true... unless Class A and Class B occur over the course of two different semesters/terms. Back in my old school it was common for some specific classes to fill up with lower-than-average students on certain semesters because of how the system was structured: Some classes you would only take in the spring semester if you'd failed it during fall or if you were behind and failed a previous class.

[thubby]

You can lose that ability without using curving as well through the differences in how difficult different classes are, or how well different teachers can put together tests that accurately measure learning the material covered. If the 90% of the class fails a test, even students that do well otherwise, it's probably either the professor's or test's fault, not the students'.

that would be a failure on the school's part, at which point they're just minimizing the damage.

if a professor wants to give hard tests, that can be done without the volatility of curving. the most obvious of which is putting more points in a test than you're grading.

[Reverent-One]

that would be a failure on the school's part, at which point they're just minimizing the damage.

*Shrugs* It's a fact of life that no one is perfect, and even if they were, getting everyone to agree that something should be done X way just isn't going to happen. Given that, minimizing the damge isn't a bad thing to do.


if a professor wants to give hard tests, that can be done without the volatility of curving. the most obvious of which is putting more points in a test than you're grading.

More points don't necessarily help with percentages. That's leaving aside the potential for the professor to simply not be great at writing tests.

[warty goblin]

Keep in mind TA classes are fairly small. And summer students I've found are almost universally better than year students. So we're talking a class of under 10 students. My largest are under 30. The probability of a class of that size actually fitting a bell curve is vanishingly small. Trying to impose a bell curve on it would amount to simply introducing a significant random element into grading.
In statistics the TA classes I've seen (at two different universities) range from 40 to 80 per section. That's more than enough to ensure decent power in a test of normality. Not that anybody should really pay that much attention to a normal test anyway. Nor have most of the grade histograms I've looked at appeared particularly normal. More like a backwards sort of gamma, really.



Actually, I considered that. It's true... unless Class A and Class B occur over the course of two different semesters/terms. Back in my old school it was common for some specific classes to fill up with lower-than-average students on certain semesters because of how the system was structured: Some classes you would only take in the spring semester if you'd failed it during fall or if you were behind and failed a previous class.

Professors tend to know about this sort of thing, and it's hardly rocket statistics to control for it.


I think the issue is that a grade ends up serving two different masters. It's supposed to measure student ability, but also professorial competence. If the test is considered infallible, there's no reason to compare one student with the others; you just set your grade categories and turn the crank. But the test is not a perfect measure of student ability. For one thing, tests have to change from year to year, so there's inter-exam variability. A 95 on this year's midterm needs to be related to a 95 on last year's midterm. Which means comparing students, did only two people score a 95 this year when ten did last year because the class sucks, or because the test is harder? If the scores in section A are lower than section B, is section A on average better, or is professor B on average worse? Or do they just write a harder test?

Because of the imperfections of the exam as a measurement tool, some level of inter-student comparison is really the only fair thing to do. Otherwise you're failing to compensate for a probable bias. Because exams are supposed to measure individual aptitude, it can't really be wholly based on comparisons. So most professors I've worked with or taken classes from seem to hammer out some sort of imprecise compromise in the middle.

[thubby]

More points don't necessarily help with percentages. That's leaving aside the potential for the professor to simply not be great at writing tests.

more points weights the things students do know.
having a test with 200 points that's graded out of 100 points means there's a solid 50% of the test you don't expect them to get.

[Gravitron5000]

I think this depends entirely on the course taken. English I would certainly agree. Engineering classes, I disagree.

I took Heat Transfer and the average on that test was about a 27%. The test scores went like this:

98, 92, 84, 69 (my score), 57 and then they dropped below that. Note this was probably a class size of about 50-60. Obviously 95% of the class failed that exam. It doesn't mean we didn't learn anything. Engineering classes are meant to be hard. Most professors understand this and compensate for it. If they were to go by the methodical grading scale of 93 or above is an A, 90 -93 A-, etc. Then a lot more people would fail engineering.

Sometimes it is just bad professors. I have had two classes where the professors taught horribly and graded harshly and failed over half the class. Still learned a lot though.

I disagree. In a field like engineering, when your lack of knowledge can cause harm, you should be tested on your knowledge of a subject, not how well you understand it relative to your peers. If you don't understand the subject matter, then you should fail. I would not want someone that barely understands shear forces to be designing bridges.

I studied engineering, and currently am employed in that field. None of my courses were graded on a curve. In my experiences in my various places of employment, I seem to be judged more on being able to meet the goals that have been set before me in a timely manner and how well I work with others than on what my co-workers are able to do. Co-operation seems to trump competition in the places I have worked.

[warty goblin]

more points weights the things students do know.
having a test with 200 points that's graded out of 100 points means there's a solid 50% of the test you don't expect them to get.

It also massively increases the grading load. This is very far from a good thing.

[Jay R]

In Differential Equations, my professor didn't like to give what he called "trivial tests". A trivial test was one that covered things you'd already had in class.

His quizzes covered what the next few weeks would be, since you now have all the background needed to solve them.

Five proofs, each worth 20 points. My first quiz came back with a grade of 21%. It was the first math test I'd ever had in which I scored less than 96%.

I spent the next 45 minutes seriously contemplating a French major.

So he showed us how each problem could be solved with the material we'd learned. Then he said, "Now about grades. Figure a 19% or more is an A."

OK - that's fair. A "non-trivial quiz" is completely reasonable if the grades are curved so that an A-level of knowledge and ability still earn an A.

[I once got one of the questions on one of his quizzes completely right. I was incredibly proud of that.]

[Emmerask]

The problem I see with this none trivial tests approach is that while its "easy" to grade those 20% students it seems almost impossible to FAIRLY grade those below that mark.
Also some might lack the needed abstraction skill to use what is learned in the course for different problems, depending on what actually is their "major subject" the abstraction skill might just not be needed, failing them then because of the lack of an unneeded skill in their future profession would be rather pointless.

Overall none standardized tests, sliding bell curves etc just decrease the significance of something already extremely hard to compare due to differently skilled professors etc ie it should not be done.

[Elder Tsofu]

And what if you got close to 100%? Would you get an instant scholarship since you're such a genius that you got 5 times the score needed for the top grade?
It is in my opinion wasted potential for distinguishing between the lower grades; to reduce the weight of chance and the whim of the one grading on the final grade.

[Haruki-kun]

If you don't understand the subject matter, then you should fail.

That's what I want to say, thank you.

Grades should be a reflection of how well you understand the subject of the class, not a quirk of academia.

[Jay R]

The problem I see with this none trivial tests approach is that while its "easy" to grade those 20% students it seems almost impossible to FAIRLY grade those below that mark.

This was a 13-person class of honors differential equations, at an elite engineering school. There weren't that many people below that mark.


Also some might lack the needed abstraction skill to use what is learned in the course for different problems, depending on what actually is their "major subject" the abstraction skill might just not be needed, failing them then because of the lack of an unneeded skill in their future profession would be rather pointless.

The only three students who weren't math majors were engineering students intending to do graduate research. Anybody not intending to do research wouldn't have taken that class. This simply didn't apply.


Overall none standardized tests, sliding bell curves etc just decrease the significance of something already extremely hard to compare due to differently skilled professors etc ie it should not be done.

Not a "differently skilled profoessor". He was Salomon Bochner (http://en.wikipedia.org/wiki/Salomon_Bochner), a highly skilled professor with an international reputation.

[Balain]

I saw something the other day. Now correct me if I am wrong. I believe they said in France everything is out of 20 and you can never get 20/20. So if I give you a test in France the only question What is 1 + 1. YOu get the correct answer I give you 19/20...weird

[Elder Tsofu]

This was a 13-person class of honors differential equations, at an elite engineering school. There weren't that many people below that mark.

Did you get an explanation for his reasoning behind taking the quite arbitrary number 19%+ as an A, or did he just chose it because most of his students didn't reach higher?

[Emmerask]

I saw something the other day. Now correct me if I am wrong. I believe they said in France everything is out of 20 and you can never get 20/20. So if I give you a test in France the only question What is 1 + 1. YOu get the correct answer I give you 19/20...weird

Duno if this really is true, however I think I can explain the reasoning.

if you just write 1+1=2 that is by no means everything you can do with this equation. You could write down the mathematical proof why it is 2.
It also could be binary with the answer 10^^.

And then tests are of course never as easy as 1+1, so if you take full marks as the complete answers to every question that could arise to the problem then getting full marks is practically impossible.

[huttj509]

Did you get an explanation for his reasoning behind taking the quite arbitrary number 19%+ as an A, or did he just chose it because most of his students didn't reach higher?

If you view the goal of the grade as "does this student know the material to the extent needed for future applications or courses" and it's a particularly difficult test, beyond the expected required proficiency level, it's not that horrid.

[Ashtagon]

In related news, http://www.bbc.co.uk/news/world-africa-23843578


Liberia students all fail university admission exam

Liberia's education minister says she finds it hard to believe that not a single candidate passed this year's university admission exam.

Nearly 25,000 school-leavers failed the test for admission to the University of Liberia, one of two state-run universities.

The students lacked enthusiasm and did not have a basic grasp of English, a university official told the BBC.

Liberia is recovering from a brutal civil war that ended a decade ago.

[Elder Tsofu]

If you view the goal of the grade as "does this student know the material to the extent needed for future applications or courses" and it's a particularly difficult test, beyond the expected required proficiency level, it's not that horrid.

I'm not horrified by the test as it is, and I don't have enough information to properly criticize (or not criticize). It is just sad that they teach bad habits through the tests since not stating requirements before grading opens up the possibility of fudging the results twice. 1) When awarding points and 2) When setting the grade requirements.
If he had wanted to he could just have handed out a note with the requirements when you hand in the test to eliminate that problem and still make the students sweat through the exam.

[warty goblin]

I'm not horrified by the test as it is, and I don't have enough information to properly criticize (or not criticize). It is just sad that they teach bad habits through the tests since not stating requirements before grading opens up the possibility of fudging the results twice. 1) When awarding points and 2) When setting the grade requirements.
If he had wanted to he could just have handed out a note with the requirements when you hand in the test to eliminate that problem and still make the students sweat through the exam.

Being able to fudge scores and requirements is absolutely a good thing.

[Jay R]

Did you get an explanation for his reasoning behind taking the quite arbitrary number 19%+ as an A, or did he just chose it because most of his students didn't reach higher?

How did you determine that 19% was arbitrary? I don't know that, and I think I have a lot more information about this 1973 class than you do.

By contrast, 90% is arbitrary. It's chosen because it's divisible by ten, and it's used in many very different courses on very different materials, measured by very different questions.

Based on everything I know about this professor, and his international reputation, and the general fairness and keen intuition he consistently displayed in class and out, that he looked over the tests, read through our attempted proofs, and determined which ones demonstrated an A level of mastery of the material, based on his (at that time) over forty years of teaching experience at the University of Munich, Princeton, and Rice.

[The tests were 5 proofs, each worth 20 points. On one later test, I got one of the questions completely correct. I was incredibly proud of that fact - far more excited than I have ever been by any other grade I've ever received. I am so glad that I was once measured with non-trivial tests.]

[Elder Tsofu]

How did you determine that 19% was arbitrary? I don't know that, and I think I have a lot more information about this 1973 class than you do.

By contrast, 90% is arbitrary. It's chosen because it's divisible by ten, and it's used in many very different courses on very different materials, measured by very different questions.

Based on everything I know about this professor, and his international reputation, and the general fairness and keen intuition he consistently displayed in class and out, that he looked over the tests, read through our attempted proofs, and determined which ones demonstrated an A level of mastery of the material, based on his (at that time) over forty years of teaching experience at the University of Munich, Princeton, and Rice.

[The tests were 5 proofs, each worth 20 points. On one later test, I got one of the questions completely correct. I was incredibly proud of that fact - far more excited than I have ever been by any other grade I've ever received. I am so glad that I was once measured with non-trivial tests.]

I did determine that it seemed arbitrary on the information provided.
Why was 19% chosen instead of 20% if only one correct proof was needed out of 5 for the top grade? Or 18%? Did he describe what a typical B was as well?

Professors generally aren't masters of everything, as we both know I'm sure. You can be an excellent teacher, mentor and/or scientist but still do some things in a way that isn't good.

[Knaight]

I did determine that it seemed arbitrary on the information provided.
Why was 19% chosen instead of 20% if only one correct proof was needed out of 5 for the top grade? Or 18%? Did he describe what a typical B was as well?

Why is 90% chosen as the generic benchmark instead of 91% or 89%. Or, for that matter, 76%. Who's bright idea was it for the entire bottom half of the score to be failing? Yes, 19% is going to be somewhat arbitrary, but so is every other system.

[warty goblin]

Why is 90% chosen as the generic benchmark instead of 91% or 89%. Or, for that matter, 76%. Who's bright idea was it for the entire bottom half of the score to be failing? Yes, 19% is going to be somewhat arbitrary, but so is every other system.

Indeed the only sensible method is to look at what was on the test, what the people taking the test knew, and what you think is an acceptable level of understanding, then calibrate accordingly.

[Elder Tsofu]

Why is 90% chosen as the generic benchmark instead of 91% or 89%. Or, for that matter, 76%. Who's bright idea was it for the entire bottom half of the score to be failing? Yes, 19% is going to be somewhat arbitrary, but so is every other system.

But the problem is that you still got a failing part in the 19% case, and if we use your generic benchmarks that would be ca. 10.6%. So 9 points out of one hundred between failing and max grade instead of 40 in your generic example. A max grade is always a max grade, so why not use the extra points to figure out where the rest of the grades happen?


Indeed the only sensible method is to look at what was on the test, what the people taking the test knew, and what you think is an acceptable level of understanding, then calibrate accordingly.

I say that it is better to turn that thinking around, write the test so they somewhat fit the students expected level of understanding instead. Unless your tests are written into stone* tablets that shouldn't be much of a hurdle.

*Or similar materials

[Jay R]

But the problem is ...

You're trying to generate a problem when the students, professor, and administration all think everything is fine.

Will you at least consider the possibility that a system that works well for everyone involved with it might not actually have a problem?


But the problem is that you still got a failing part in the 19% case, and if we use your generic benchmarks that would be ca. 10.6%. So 9 points out of one hundred between failing and max grade instead of 40 in your generic example. A max grade is always a max grade, so why not use the extra points to figure out where the rest of the grades happen?

I repeat - Honors differential equations at an elite private school. There was nobody in the class who was likely to fail. I doubt if there was anyone in the class who ever made less than an A in a math class before arriving at Rice University.


I say that it is better to turn that thinking around, write the test so they somewhat fit the students expected level of understanding instead. Unless your tests are written into stone* tablets that shouldn't be much of a hurdle.

*Or similar materials

"somewhat fit the students expected level of understanding"? That's exactly what it did. These were top-level math students. The tests were written to give us the experience of trying to prove theorems we'd never seen before, extending our own knowledge. It was a great challenge for us, unlike any other math test I'd ever taken. That course was a tremendous help to me on my way to my own Ph.D.

I never heard a single student complain about the tests, though we all talked about how unusual they were.

[warty goblin]

But the problem is that you still got a failing part in the 19% case, and if we use your generic benchmarks that would be ca. 10.6%. So 9 points out of one hundred between failing and max grade instead of 40 in your generic example. A max grade is always a max grade, so why not use the extra points to figure out where the rest of the grades happen?

Assuming you use pluses and minuses, and that C- is the last passing grade, there's only seven grades between A and D+. You don't need forty points to work that out. Statistically speaking, the wider range of points in fact decreases the meaningful precision of the test, since it will result in larger standard errors and thus increase the size of the smallest difference that can be considered significant. Put slightly differently, over a forty point spread there probably isn't any significant difference between two people who's scores differ by a point, yet they can very easily receive quite different grades. Over a nine point spread though? Much more likely to actually indicate different levels of competence.

(The downside of a small spread is that it magnifies the effect of grading error. For a small, high level class however grading error is probably both small, and likely to be caught by the students.)

(I'm implicitly using a model where score = true competence + student error + grader error, where both student and grader error are random, and errors in the statistical sense of the word. Grade is of course just a discrete step function of score.)



I say that it is better to turn that thinking around, write the test so they somewhat fit the students expected level of understanding instead. Unless your tests are written into stone* tablets that shouldn't be much of a hurdle.

*Or similar materials
This is the general practice. However expected and actual do not necessarily match up, so post-test adjustments are a good option to have on the table. Testing is an art, however much people want it to be a science.

[Dr Mushroom]

I came into this thinking you meant some kind of homebrewed tabletop gaming class.

[Knaight]

Indeed the only sensible method is to look at what was on the test, what the people taking the test knew, and what you think is an acceptable level of understanding, then calibrate accordingly.

I would also consider calibrating the test to an existing scale acceptable, though it's not necessarily a great way to handle it. So is having tests be something that drags everyone down, while other parts of the class counterbalance that.

With that said, that leaves a lot of stupid methods. My personal least favorite that I've seen is setting 100% at whatever the best student in the section got (though this was more of a high school thing). I've seen that either cause issues with outliers, or have the potential to with teachers who used a better system, and this includes cases where I was that outlier.

[warty goblin]

With that said, that leaves a lot of stupid methods. My personal least favorite that I've seen is setting 100% at whatever the best student in the section got (though this was more of a high school thing). I've seen that either cause issues with outliers, or have the potential to with teachers who used a better system, and this includes cases where I was that outlier.
I've had the max = 100% system used. Worked very well for computer science, but I think it would probably do very badly in a lot of other classes.

[Emmerask]

Problem with the best test == 100% system are students who have prior knowledge in the subject.
For example I was a java dev for over a year before I had to take programming 1 at university (which was java).

If the test would then have been rated using me as the 100% mark then it in no way would have reflected how much was learned. (I was present 3 times during the semester including the test day :smallwink:)

So maybe take the (depending on class size) x best tests, avg them and then use them as 100%?

[Knaight]

I've had the max = 100% system used. Worked very well for computer science, but I think it would probably do very badly in a lot of other classes.

It does, particularly in smaller classes. I've kicked everyone else down a letter grade on the final before (Physics), and had my Econ teacher curved based on a student score instead of what he thought was an A-level I'd have kicked everyone who didn't exceed the 100% point down 2 - and these are just the classes where I was the outlier. I've also taken classes where one of the students was truly brilliant (to the point where I wouldn't be surprised to see a Nobel with their name on it in the future), which had pretty hard tests. The score breakdowns routinely looked something along the lines of 100%, 70%, 60%, 59%, 58%, 57%, etc, and that scaling would have adversely affected a lot of grades unfairly, along with being a recipe to manufacturing hostility towards the brilliant student.

[Jay R]

I've had the max = 100% system used. Worked very well for computer science, but I think it would probably do very badly in a lot of other classes.

The problem with it is that if one person (who made the top grade) were not in the class, everybody's grade gets better. If one person who could make a higher grade joined the class, everybody's grade gets worse.

If your grade depends on who else happens to be in the class, rather than on your own work and understanding alone, then it's arbitrary.

I have no problem with a teacher looking at a test and deciding that, for instance, a 65% should be a B. If the level of knowledge demonstrated by a 65 represents a B level of understanding of the material, that's fine. But I object strongly if that B becomes a D because a single student made a 97 on the test.

Once in third grade, I had to sneak out of the school from a different entrance, and go home an unusual way, because every other student in the class knew that they lost two letter grades just because I did well on the test.

[warty goblin]

The problem with it is that if one person (who made the top grade) were not in the class, everybody's grade gets better. If one person who could make a higher grade joined the class, everybody's grade gets worse.

If your grade depends on who else happens to be in the class, rather than on your own work and understanding alone, then it's arbitrary.

I have no problem with a teacher looking at a test and deciding that, for instance, a 65% should be a B. If the level of knowledge demonstrated by a 65 represents a B level of understanding of the material, that's fine. But I object strongly if that B becomes a D because a single student made a 97 on the test.

Once in third grade, I had to sneak out of the school from a different entrance, and go home an unusual way, because every other student in the class knew that they lost two letter grades just because I did well on the test.

The grade is arbitrary one way or the other. All setting max = 100% does is set the maximum. It doesn't determine where the bottom of the As is, only that at least one person is getting an A. And in my experience it worked just fine, particularly in a class as funky to grade as CS.

I don't, as a rule, get particularly fussed about grades. I just do the best work I can; what the prof decides to call it is up to them. If they use a system that compares me to other students, that's fine. If they don't, that's also fine. The only times I've ever felt I've been unfairly graded are when I'm held responsible for failing a test that upwards of fifty percent of the class failed. These were graduate level math classes, so it wasn't that half the class didn't care, it was that the professor couldn't find their ass with both hands and a map. To my mind the only fair thing to do in those circumstances is admit that the test itself is badly designed and scrap the whole thing, because it's far more likely than three quarters of the class being incompetent and unmotivated.

Not that anybody ever does that of course. If the choice is between a professor admitting they have failed in a stupid way, and making other people suffer a bit, get ready to suffer.

[thubby]

The grade is arbitrary one way or the other. All setting max = 100% does is set the maximum. It doesn't determine where the bottom of the As is, only that at least one person is getting an A. And in my experience it worked just fine, particularly in a class as funky to grade as CS.

setting a maximum is perfectly reasonable since any given class is only designed to teach certain things.

[Jay R]

Not that anybody ever does that of course. If the choice is between a professor admitting they have failed in a stupid way, and making other people suffer a bit, get ready to suffer.

Not always. When I give a test, I start taking it at the same time my students do. Sometimes I find a typo that makes it harder than it was intended to be, in which case I stand up and say, "Attention please. Question 3 has a typo. Use this equation (write on board) instead of the one on the test.

If an hour-long test takes me longer than 6 minutes to complete, I'll tell the class to ignore one or two questions.

[Knaight]

The grade is arbitrary one way or the other. All setting max = 100% does is set the maximum. It doesn't determine where the bottom of the As is, only that at least one person is getting an A. And in my experience it worked just fine, particularly in a class as funky to grade as CS.

From what I've seen, setting max=100% usually also involves setting the rest of the scale, along with the use of the standard [90-100%] A, [80-90%) B, etc. The highest student getting a 90 on a 100 point test places everyone in the 81 to 90 range as an A. If that aspect of the system is removed there ceases to be a problem, but with it there the max=100% system breaks horribly as soon as an outlier on the high end shows up in a class.

[Jay R]

The problem is that the highest score is an outlier, and therefore a very volatile point. Making everybody's grade dependent on who happens to be best in this class gives an unfairly low grade to the people in the section with one genius, and an unfairly high measurement of the people in the class with no A-level students.

"Your grade in this course will be based on how well that person over there learns the material. If that person is really great, then your score will go down. If that person does poorly, then your score will go up."

Leaving aside the rampant unfairness of basing your score on somebody else's, that mean that the best way to make a good grade isn't to learn the material, but to blackmail the best student in the class to deliberately give wrong answers. (And yes, back in elementary school, I had classmates attempt to do exactly that.)

My grade should be based on my work, and nobody else's. And also, other people's grade should not be based on my work, but only on their own.

[warty goblin]

The problem is that the highest score is an outlier, and therefore a very volatile point. Making everybody's grade dependent on who happens to be best in this class gives an unfairly low grade to the people in the section with one genius, and an unfairly high measurement of the people in the class with no A-level students.

"Your grade in this course will be based on how well that person over there learns the material. If that person is really great, then your score will go down. If that person does poorly, then your score will go up."

Leaving aside the rampant unfairness of basing your score on somebody else's, that mean that the best way to make a good grade isn't to learn the material, but to blackmail the best student in the class to deliberately give wrong answers. (And yes, back in elementary school, I had classmates attempt to do exactly that.)

My grade should be based on my work, and nobody else's. And also, other people's grade should not be based on my work, but only on their own.

If the best way to score well in this system is blackmail, the best way to score well with a conventional test is pictures of the professor in a compromising position with livestock dressed as cheerleaders. Which, alas, I've never been able to procure.

On a more serious note, all I can say is that it worked very well in the classes I had which used it. About half the class was usually some variety of incompetent, and the other half quite good, so the incompetent people still took their lumps, and the competent people didn't get punished because the professor accidentally wrote a really hard problem every now and again. Tests are fallible instruments after all, I've taken enough of them to have observed this first hand. I'd much rather have some smidgen of my grade determined by other students than the entire thing by the professor. At least I know some members of the class are trying, and I've definitely had some professors who were phoning it in.

Setting max = 100% doesn't penalize me for one genius who writes a super-good test any more than a conventional out of 100 rubric does. If they write close to 100%, and I suck, I'll get a sucky grade regardless of whether I'm graded out of the full hundred points, or the top score in the high nineties. Actually under this system I'm probably better off, unless the best person in the class writes a perfect test in which case I'm in the boat you're advocating for anyway. All the system does is calibrate the scale, which I appreciate because tests are very fallible instruments.