Bell curve system (1 Viewer)

[Cyph]

UNSW uses a bell curve system to grade their students...

The bell curve is normally distributed data isn't it So they'd apply 3 tandard deviations, giving 6 different 'groups', which could be broken down into from R-L, HD, D, C (two middle SDs together), P, F (graph here in red & green if a picture helps - http://www.robertniles.com/stats/stdev.shtml ). Or it could be from R-L, HD, D, C, P, CP, F

If there's 93 students in the course, this means 5% (5 students after rounding up) get a Distinction...

I don't think this seems right, so I'm thinking I could be interpretting how the bell curve would be used incorrectly?

Does anyone know for sure how it's done?

 

[...]

flyin' could give you a really good intepretion..he made a post somewhere..let me find it

 

[Cyph]

I want to know *exactly* how grades are worked out.

 

[Exeter]

really?
my marketing lecturer said he had no limit on the number of hd's he was allowed to give out

 

[t-i-m-m-y]

i gues it depends

for BIOS1201. its just raw marks. no scaling

 

[Lazarus]

Originally posted by Cyph
UNSW uses a bell curve system to grade their students...

Most faculties within UNSW do, but not all. The exact parameters of the bell curve used can also vary between faculties.

The bell curve is normally distributed data isn't it

Yes.

So they'd apply 3 tandard deviations, giving 6 different 'groups', which could be broken down into from R-L, HD, D, C (two middle SDs together), P, F. Or it could be from R-L, HD, D, C, P, CP, F

I suppose that's one way of looking at it... though I'm not sure it would really work out that way (using the magnitude of one standard deviation for each grade).

If there's 93 students in the course, this means 5% (5 students after rounding up) get a Distinction...

No. It depends on the parameters of the curve; the data could be normally distributed with a mean of 50 and a standard deviation of 29 (which would tend to spread marks over the full range), or the data could be normally distributed around 68 with a standard deviation of 12 (closer to a university distribution).

In general, though, universities don't strictly apply any sort of bell curving formula. As far as I'm aware, most faculties use a multi-linear mapping procedure... raw cut-offs are determined for each grade such that the desired proportions are (roughly) obtained. (Similar to the 'aligning' system used by the Board.) Those desired proportions are generally similar to the ones given by a bell curve. But the mapping procedure is much more flexible than a bell curving one (allowing more high grades where a candidature is particularly good etc). Lecturers are allowed to depart from the bell curve distribution (not sure whether they actually do)... they just have to be able to justify it with reference to the candidature.

 

[hk31]

i think we were told in the first week of session that they don't use the bell curve system for commerce subjects....

 

[Minai]

Originally posted by hk31
i think we were told in the first week of session that they don't use the bell curve system for commerce subjects....

Whoever told you that was lying.

 

[Cyph]

cool, thanks a lot Laz!

They wouldn't be bell curving marks after every assessment would they? They'd wait until they got the overall raw final marks, then bell curve them?

 

[Survivor39]

Originally posted by t-i-m-m-y
i gues it depends

for BIOS1201. its just raw marks. no scaling

Then how the hell someone got 98 last year? Unless he/she is a super nerd.

It's weird because Paul adam said the top mark for last year was 94, then my demo said it's 98....

 

[Lazarus]

Originally posted by Cyph
They wouldn't be bell curving marks after every assessment would they? They'd wait until they got the overall raw final marks, then bell curve them?

Mmm... again, it varies. My experience with UNSW (which is limited to CSE and Law) is that the marks are adjusted for each task. The sum of your marks is your final mark. However, when I was at USYD, this wasn't the case, and your final mark could actually differ significantly from the sum of your marks.

 

[McLake]

Peer assesment can also contribute to scaling (if your course has it)

 

[googy_1985]

Down here at Canberra (UC) they havent used any scaling for about 5 years now . The uni etc arent allowed to scale, we are given what makr we deserve, BUT all lecturers have to submit their assessments and marking criteria to the divisional board or something to ensure it is at the right standard.

 

[redruM]

~68% will lie within 1 sd, ~96 will lie within 2 sd of the mean..for normally distributed data.

so that would suggest:
2% will get HD,
the next 14% will get D,
next 68% credit,
next 14% pass,
and 2% fail

or
2% will get HD,
the next 14% will get D,
next 34% credit,
next 34% conceded pass
next 14% pass,
and 2% fail

thats based on the assumption that was stated...

Originally posted by Cyph
The bell curve is normally distributed data isn't it So they'd apply 3 tandard deviations, giving 6 different 'groups', which could be broken down into from R-L, HD, D, C (two middle SDs together), P, F (graph here in red & green if a picture helps - http://www.robertniles.com/stats/stdev.shtml ). Or it could be from R-L, HD, D, C, P, CP, F

i would have thought they worked on the outlier model, where anything 1.5 sd's away from the mean was excetional, exceptionally poor[F] or exceptionally good[D or HD]. then C and P's would fit within the 86% within 1.5 sd's of the mean?

 

[Cyph]

redruM, interesting post.. thanks!

I guess all we can do is speculate

 

[redruM]

Originally posted by Cyph
redruM, interesting post.. thanks!

lol, no worries. this is somehow substituting for my study for stat. lol.

everyone is given the same treatment, so whatever the system, we can be sure it is fair, unless you fail, of course

 

[hk31]

Originally posted by Minai
Whoever told you that was lying.

oh "great"...that's just fantastic...

 

[gman03]

So that means if I enrolled in the same class with a bunch of smart people and everyone beat me in every assessment then I will fail no matter how hard I try? i.e. the bottom 2% of the S.D. curve?

That is just worse than the HSC.

 

[Minai]

Originally posted by hk31
oh "great"...that's just fantastic...

Yeah =/
For Commerce subjects, the mark you get for your assessments (not finals) are THE marks you actually get. However, after you do the final, your marks are then bell-curved/standardised etc
ECON1101 for example, I was on about 27/40 after all the assessments, then after the final exam (worth 60%) I got 85 overall, meaning, if things were not standardised, I wouldve got 97% in the final exam which could not really be possible.

I could be wrong, but my tutors have indicated this, and thats just the impression I get from personal experience

 

[sunny]

Originally posted by gman03
So that means if I enrolled in the same class with a bunch of smart people and everyone beat me in every assessment then I will fail no matter how hard I try? i.e. the bottom 2% of the S.D. curve?

That is just worse than the HSC.

ehhh....if you've read the thread, thats not the case everywhere.