I use an analytic scoring scheme. Students start at 100% and I lose points for mistakes with my approach. This is true for both in-class and take-home exams. I give partial credit for answers, I tend to categorize the number of points for each part of a question. For example, if I ask a hypothesis test question for a proportion, I would have points for the correct hypotheses, the correct test statistic, for the P-value, the interpretation of the result and the check for conditions. I do not have an expected average for each test. I do not curve the exam scores. Generally I have at least a couple students score in the high 90's, so I do not curve. In a rare situation where I have given an exam that was indeed too long, I will do some kind of curve, but that does not happen very often. I use a straight total at the end of the semester to determine the final grade. The letter grades of earlier tests are not important, it is the final weighted average that I use to assign final grades. I do not curve the final percentages. I find that I have a large number of "A" grades, so I am not pressured by anyone to raise grades. The big lurking variable is the 20% homework grade. I allow students to resubmit their group homework for a higher grade and so almost all students end up with a homework grade in the 90's. This acts as my "curve". Thus I do not feel bad about giving a student with an 89.3 average a B+ or a student with a 59.2 an F.

See all answers to this question

See all of John's answers