So. I just graded the final exams for our applied regression class. Lots of students made mistakes which gave me the feeling that I didn’t teach the material so well. So I thought it could help lots of people out there if I were to share the questions, solutions, and common errors.
It was an in-class exam with 15 questions. I’ll post the questions and solutions, one at a time, for the first half of June, following the model of my final exam for Design and Analysis of Sample Surveys from a few years ago. Enjoy.
To help me figure out if my exam questions or teaching are problematic, I find it useful to plot for each exam I give the correlation of each question’s score with the overall exam score vs. the fraction of students answering that question correctly. The bottom left quadrant shows questions that many people get wrong *and* that are not well correlated with overall exam performance. These are good candidates to be tossed out of the grading, implying bad questions, a mismatch between what I taught and what I thought I taught, or both. An example and more text is here. (I could give more examples, in which the bottom left is more populated!)
There’s a term from the educational literature for this exam question validity analysis based on correlation between questions, but I can’t remember what it is!
Raghuveer:
It’s called item-response modeling, and you’re looking for questions with poor discrimination.
Thanks!
“first half of June”? That’s quite a tease!
I’m guessing that he is referring to the time he writes the posts (so if what he wrote in the beginning of June is appearing now, that would imply that the queue has about a 6 1/2 month delay).
Martha:
Actually, I wrote the post just now, and I put the posts for the 15 questions and answers in June, 2019, which is where the queue is now.
I stand corrected — so Dave is right: quit a tease.
Do you do any curving of grades?
http://www.statisticool.com/curvedgrades.htm (or other methods?)
Justin
http://www.statisticool.com