A message to all subject coordinators
Most universities expect you to end up with a distribution something along the lines of
High HD/H1 (90-99%) : a few exceptional students
Normal HD/H1 (80-90%): A small but significant percentage of the class
The bulk of the students that you want passing distributed through the rest of the passing grades
Almost no-one in the 40% to 50% range **
Slackers who know something: 30% to 40%
Never attended a class: 0% to 10%
**The reason you don't want students in the 40% to 50% range is that it's just painful for everyone. It doesn't make them say "Oh, if I'd just worked a leeeettle bit harder, I'd have passed!" and then work harder next semester. It makes them bitter and pissed off that something small has fucked up (or robbed them of) a year of their life. Consequently, a lot of subject coordinators - in my eyes, generally the good ones - try to avoid giving marks that end in 9s
These mark distributions are fairly standard. They're not that hard to achieve. Most courses have been around for quite a while, and there's a lot of information about how well students know particular areas, what they find hard...
If you're running multiple courses over the same area (eg: first year advanced maths, standard first year maths, first year maths for those who need a hand up), yes, some form of scaling might be appropriate.
For all other classes - don't screw around with scaling. It just makes things distressing and awkward for everyone.
Designing an exam to achieve the distribution above IS NOT THAT FUCKING HARD. If you have many of the better students in your class coming out of the exam shaking their heads, counting their marks and expecting poor results, you have fucked up. If you then scale everything to get the appropriate distribution (like the one above), you have not solved the problem, you have addressed the symptom, and you are still shit at what you do.
If you are a junior staff member, get senior colleagues to look at your work. Exams are important to students.
So, how does one get the appropriate distribution?
(1) Cover topics with weights resembling the amount of time spent on each area during the course
(2) Ask questions on each topic with a similar difficulty to the practice questions encountered during the course.
(3) Disambiguating questions - those that distinguish the cream from the norm - should not constitute a larger portion of the marks than the amount by which you want the high achievers to be distinguished
(4) Norm-boosters or "gimmes" should cover topics that a reasonable acquaintance with the subject should make simple, but inexperience should make very difficult
Further to (2) Ask questions of moderate difficulty. Ask predominantly moderately difficult questions. They create a nice, shiny, normal distribution by simple virtue of common mistakes. Asking predominantly gimmes and doozies fucks up your distribution.
Further to (4): Just because a topic is "hard" is no reason to cover that topic only with "gimmes". Moreover, it can be counter-productive. If the topic is known to be hard, then "gimmes" are more likely to be even simpler, and can be answered by someone off the street. If questions on any topic are too simple, a moderate student is likely to interpret them as a "trick question".
A good example of a "gimme" for pumps:
"What are the effects on pressure-head gain and system flow of combining pumps in parallel?"
A bad example:
"Add these two very difficult to derive, but very simple equations"
Make the easy questions require some background. For the bulk of topics, ask some basic background questions AND ask some questions that require some working, that familiarity will make simple, but inexperience will make awkward AND OPTIONALLY some tricky "bonus" questions that allow the swots to shine.
Or, if you're only asking a couple of long questions, allow each question to show students (1) know what's involved (2) know how to deal with it (3) understand extensions and implications of the topic.
It's not that fucking hard.
High HD/H1 (90-99%) : a few exceptional students
Normal HD/H1 (80-90%): A small but significant percentage of the class
The bulk of the students that you want passing distributed through the rest of the passing grades
Almost no-one in the 40% to 50% range **
Slackers who know something: 30% to 40%
Never attended a class: 0% to 10%
**The reason you don't want students in the 40% to 50% range is that it's just painful for everyone. It doesn't make them say "Oh, if I'd just worked a leeeettle bit harder, I'd have passed!" and then work harder next semester. It makes them bitter and pissed off that something small has fucked up (or robbed them of) a year of their life. Consequently, a lot of subject coordinators - in my eyes, generally the good ones - try to avoid giving marks that end in 9s
These mark distributions are fairly standard. They're not that hard to achieve. Most courses have been around for quite a while, and there's a lot of information about how well students know particular areas, what they find hard...
If you're running multiple courses over the same area (eg: first year advanced maths, standard first year maths, first year maths for those who need a hand up), yes, some form of scaling might be appropriate.
For all other classes - don't screw around with scaling. It just makes things distressing and awkward for everyone.
Designing an exam to achieve the distribution above IS NOT THAT FUCKING HARD. If you have many of the better students in your class coming out of the exam shaking their heads, counting their marks and expecting poor results, you have fucked up. If you then scale everything to get the appropriate distribution (like the one above), you have not solved the problem, you have addressed the symptom, and you are still shit at what you do.
If you are a junior staff member, get senior colleagues to look at your work. Exams are important to students.
So, how does one get the appropriate distribution?
(1) Cover topics with weights resembling the amount of time spent on each area during the course
(2) Ask questions on each topic with a similar difficulty to the practice questions encountered during the course.
(3) Disambiguating questions - those that distinguish the cream from the norm - should not constitute a larger portion of the marks than the amount by which you want the high achievers to be distinguished
(4) Norm-boosters or "gimmes" should cover topics that a reasonable acquaintance with the subject should make simple, but inexperience should make very difficult
Further to (2) Ask questions of moderate difficulty. Ask predominantly moderately difficult questions. They create a nice, shiny, normal distribution by simple virtue of common mistakes. Asking predominantly gimmes and doozies fucks up your distribution.
Further to (4): Just because a topic is "hard" is no reason to cover that topic only with "gimmes". Moreover, it can be counter-productive. If the topic is known to be hard, then "gimmes" are more likely to be even simpler, and can be answered by someone off the street. If questions on any topic are too simple, a moderate student is likely to interpret them as a "trick question".
A good example of a "gimme" for pumps:
"What are the effects on pressure-head gain and system flow of combining pumps in parallel?"
A bad example:
"Add these two very difficult to derive, but very simple equations"
Make the easy questions require some background. For the bulk of topics, ask some basic background questions AND ask some questions that require some working, that familiarity will make simple, but inexperience will make awkward AND OPTIONALLY some tricky "bonus" questions that allow the swots to shine.
Or, if you're only asking a couple of long questions, allow each question to show students (1) know what's involved (2) know how to deal with it (3) understand extensions and implications of the topic.
It's not that fucking hard.