4 ms·

I have never understood scaling, exams are meant to provide employers and yourself with a rating of how well you know the subject, ideally in comparison to ever

I have never understood scaling, exams are meant to provide employers and yourself with a rating of how well you know the subject, ideally in comparison to everyone else in the field. So if you fail, but everyone else fails too, you get an A. If yo do 'well' but a group of 5 out of 30 excel, you get a lower than representative grade (representative of your knowledge). How about making the questions worth points and the 100 points = 100%? 93+ = A, 80-93 = B etc. That's how all 3 of my undergrad degrees worked. I knew what I needed to do for a grade and damn everyone else.

aninhumer 11y ago
Normalisation makes more sense over larger populations. If most people across a large group do much worse on a particular exam, is it more likely that everyone screwed up, or that the exam was unusually hard?
- kagamine 11y ago
  Universities I have attended have procedures to deal with those instances in which everyone performs equally badly. It is almost always an indication that something else is amiss. Your answer has not convinced me that giving an A to a student who got 55% correct on a test displays a command of the subject simply because everyone else got 30% correct.
  - aninhumer 11y ago
    Sure, in exceptional cases investigation is probably necessary. But there are still more marginal cases where one exam gets worse results than another. >Your answer has not convinced me that giving an A to a student who got 55% correct on a test displays a command of the subject simply because everyone else got 30% correct. The point is that it's just as unconvincing to say the student that got 55% is comparable to another student that got 55% in an exam where that was the lowest score. Exams are not perfectly comparable measures of absolute "mastery", so we have to use statistics to try and derive as much information as we can from their results. If students in one cohort consistently perform worse, it's probably their fault, but if students in one exam perform worse, it's more likely the examiner's fault.
cozzyd 11y ago
A lot of professors in my field (physics) see tests as a learning tool where one is presented with new things to be figured out. In my undergrad, exams were typically open-book and consisted of a few long, multipart problems. It does not make sense to scale such exams to 100%. Anyway, by setting the expected mean to around 50%, you get a more accurate evaluation of people's understanding (by minimizing truncation at the limits).

aninhumer 11y ago

Normalisation makes more sense over larger populations. If most people across a large group do much worse on a particular exam, is it more likely that everyone screwed up, or that the exam was unusually hard?

kagamine 11y ago
Universities I have attended have procedures to deal with those instances in which everyone performs equally badly. It is almost always an indication that something else is amiss. Your answer has not convinced me that giving an A to a student who got 55% correct on a test displays a command of the subject simply because everyone else got 30% correct.
- aninhumer 11y ago
  Sure, in exceptional cases investigation is probably necessary. But there are still more marginal cases where one exam gets worse results than another. >Your answer has not convinced me that giving an A to a student who got 55% correct on a test displays a command of the subject simply because everyone else got 30% correct. The point is that it's just as unconvincing to say the student that got 55% is comparable to another student that got 55% in an exam where that was the lowest score. Exams are not perfectly comparable measures of absolute "mastery", so we have to use statistics to try and derive as much information as we can from their results. If students in one cohort consistently perform worse, it's probably their fault, but if students in one exam perform worse, it's more likely the examiner's fault.

cozzyd 11y ago

A lot of professors in my field (physics) see tests as a learning tool where one is presented with new things to be figured out. In my undergrad, exams were typically open-book and consisted of a few long, multipart problems. It does not make sense to scale such exams to 100%. Anyway, by setting the expected mean to around 50%, you get a more accurate evaluation of people's understanding (by minimizing truncation at the limits).