A class of 200 students performing much worse than the last class is very unlikely. 200 Students is enough to make even small differences statistically significant.
A single test being much harder than the last test is much more likely, since it isn’t an averahe of 200, it’s a single datapoint.
That’s why if this semester’s class performed much worse than last semester’s, you can assume it’s because of the test, not the students.
Not unlikely enough, even something as simple as more students studying harder between years would mean the next set of average students drop in score despite the same performance.
Grading on a curve is always unfair when the grade carries forward and isn’t just for a one-off application. More unfair when classes are smaller and student cohorts differ.
I guess this makes sense in a university context and with aggregated data.
As an example of how you do not want to do grading based on an average, I once had a high school professor rescale my 85%-ish percent on a test to 65%-ish, because most people did well in that test so the professor decided he had made the test too easy and scaled grades down.
That was only one of the reasons I hated that guy’s guts.
I’m okay with scaling grades up because that implies either the test or instruction was bad, and the curve accounts for that. Going the other way unfairly punishes things like misreading questions.
Or when there are 1000 students over multiple classes getting 5 different versions of the test (to make looking over someone’s shoulder more difficult.)
If one of those has a significantly lower average it’s more likely it just had a few badly worded questions than that those 200 randomly picked students are all bad at the given subject.
A class of 200 students performing much worse than the last class is very unlikely. 200 Students is enough to make even small differences statistically significant.
A single test being much harder than the last test is much more likely, since it isn’t an averahe of 200, it’s a single datapoint.
That’s why if this semester’s class performed much worse than last semester’s, you can assume it’s because of the test, not the students.
Or more likely, instruction was lacking and didn’t cover the material on the test properly.
Not unlikely enough, even something as simple as more students studying harder between years would mean the next set of average students drop in score despite the same performance.
Grading on a curve is always unfair when the grade carries forward and isn’t just for a one-off application. More unfair when classes are smaller and student cohorts differ.
I guess this makes sense in a university context and with aggregated data.
As an example of how you do not want to do grading based on an average, I once had a high school professor rescale my 85%-ish percent on a test to 65%-ish, because most people did well in that test so the professor decided he had made the test too easy and scaled grades down.
That was only one of the reasons I hated that guy’s guts.
Yeah, that’s unfair.
I’m okay with scaling grades up because that implies either the test or instruction was bad, and the curve accounts for that. Going the other way unfairly punishes things like misreading questions.
Or when there are 1000 students over multiple classes getting 5 different versions of the test (to make looking over someone’s shoulder more difficult.)
If one of those has a significantly lower average it’s more likely it just had a few badly worded questions than that those 200 randomly picked students are all bad at the given subject.