In Quiz too long and too hard, I reported on the first quiz of the circuits course, and how I had made it much too tricky for the course.
Yesterday I gave another quiz covering the material of the first quiz (impedance, voltage dividers, low-pass and high-pass RC filters) plus some new material on inverting and non-inverting amplifiers using op amps. I also asked a fairly big design question that ended up being worth about 40% of the points:
Design a circuit that takes an AC input signal with a frequency above 1Hz and below 50kHz that is centered at about 1v and has a swing of ±100mV, and provides an in-phase AC output signal that is centered at 2.5V and has a swing of ±1V. (Note: “in-phase” means that the signals move up and down together.) Draw a block diagram of your design and a schematic. You may assume that you have an external 5V power supply.
I was aiming for a test that would have a mean of 50% and a standard deviation of 20%, for a “typical” class. (Of course, given that this is the first time I’ve taught the class, I still have no calibration for what a typical class should look like, so I’m basing my estimates on where I think the current class should be, given what we have covered). I knew from taking the quiz at the same time as the students that it was a bit too long. It took me 26 minutes to do all the problems, and they had 70 minutes. I usually figure that I should be able to do their 70-minute quizzes or exams in 20 minutes, so the quiz was about 30% too long. The group tutor for the class also took the quiz, and I graded both his quiz and mine before grading the class. I gave myself 95% (I’d done a divide by 2π instead of multiply by 2π in one place, and my schematic was missing some labels and a capacitor to keep the virtual ground low-noise). I gave the group tutor 65%, which was a little lower than I expected from him (though, of course, he has not been studying for this course, but just relying on what he has learned in previous courses). His score was still higher than anyone taking the class got.
I just finished grading the quiz, and the mean was 38.55 with a (sample) standard deviation of 14.6 (out of 100 points), range 13–60, median 35. I had to be fairly generous in my partial credit for the design problem, looking for pieces of the circuit that made sense, as no one got the entire design right. A lot of the block diagrams were nonsense, looking more like random collections of ideas than a block diagram (the people who teach “mind mapping” have a lot to answer for). If we correct for the quiz being 30% too long, it looks like the difficulty level was about where I intended it to be, at least with the generous grading.
I’ll have to assign letter grades to everyone on this quiz (I didn’t for the first quiz—since I decided not to count that quiz after I saw how poorly everyone did on it), and I’m not sure where to make the cut points. The highest scorer in the class was still only what I’d consider a B+ performance—the design problem was not bad (about 2/3 credit), but there were a lot of errors in the earlier, more straight-forward problems. At the low end, I need to decide whether then 13 is a low pass or not passing, and what grade to give the second lowest score—it was clearly passing, but at what level?
[Correction 2013 Mar 3: I regraded some of the quizzes, looking to see if there was any partial credit I had overlooked on the first pass. This made a big difference at the bottom of the distribution, changing the range to 22–60, µ=39.45, σ=13.1. With those changes, everyone passes.]
I had graded the quizzes a problem at a time, not looking at the names of the students, so I was curious to see how students ranked compared to how they’ve been doing on the lab reports. As it turned out, I was not surprised by the identities of the top and bottom scorers, but some of the other ranks surprised me a bit, with one of the students whose lab reports were generally poor being near the top of the class on the quiz, and one whose lab reports where generally fairly good near the bottom.