Constructing Exams In constructing exam questions, do you focus on the content and construct a realistic setting for the questions or do you start with a real problem and try to match the course content? In other words, how much do you strive to use real data as opposed to hypothetical or realistic data in the exam questions?

In constructing the exam, how do you decide how many points a question should be worth? What percentage of the points on the exam would you say are for primarily conceptual knowledge/interpretation vs. calculation/mechanics?

When writing an exam, I think it is important that each question have a context, since without a context, it seems that interpretation is not very meaningful. However, I do start with content. I usually make a list of the type of questions I want to include, and then try to write questions with a reasonable context to fit the content questions that I want to ask. I am a real advocate of real data in texts, homework, and especially in any kind of data analysis type project work, but for exams I usually settle for realistic because it usually doesn't take as much in the way of explanation to set them up. I don't want to have students to have to spend an inordinate amount of time reading background information in a timed exam, but I do want them to have a context so that I can see if they can provide conclusions in context. I would take a completely different approach if I were making up a take-home exam.

I would guess that about 30-40% of my exam points are calculation/mechanics, and 60-70% on interpretation and conceptual understanding.