Using group comparisons for grading is appropriate when the class size is sufficiently large (perhaps 35 students or more) to provide a reference group representative of students typically enrolled in the course. The following steps describe a widely-used and generally sound procedure:
- Convert raw scores on each exam to a standard score (z or T) by using the mean and standard deviations from each respective test, set of papers, or presentations. Standard scores are recommended because they allow us to measure performance on each grading component with an identical or standard yardstick. When relative comparisons are to be made, it is not advisable to convert raw scores to grades and average the separate grades. This is because the distinction between achievement levels will be lost; differences will melt together as students are forced into a few broad categories.
- Weight each grading variable before combining the standard scores. For example, double both exam standard scores and the standard score for the paper, triple the final exam standard score, and do nothing to the standard score for the presentation. The respective weights for these variables in the total will then be 20 percent, 20 percent, 20 percent, 30 percent, and 10 percent.
- Add these weighted scores to get a composite or total score.
- Build a frequency distribution of the total scores by listing all obtainable scores and the number of students receiving each. Calculate the mean, median, and standard deviation. Most calculators now available will perform these operations quickly.
- If the mean and median are similar in value, use the mean for further computations. Otherwise use the median. Let's assume we have chosen the median. Add one half of the standard deviation to the median and subtract the same value from the median. These are the cutoff points for the range of C's.
- Add one standard deviation to the upper cutoff of the C's to find the A- B cutoff. Subtract the same value from the lower cutoff of the C's to find the D-F cutoff.
- Use number of assignments complete or quality of assignments or other relative achievement data available to reevaluate borderline cases. Measurement error exists in composite scores too!
Instructors will need to decide logically on the values to be used for finding grade cutoffs (one-half, one-third, or three-fourths of a standard deviation, for example). How the current class compares to past classes in ability should be judged in setting standards. When B rather than C is considered the average grade, step five will identify the A-B and C-B cutoffs. Step six would be changed accordingly.
Relative grading methods like the one outlined above are not free from limitations; subjectivity enters into several aspects of the process. But a systematic approach similar to this one, and one which is thoroughly described in the first class meeting, is not likely to be subject to charges of capricious grading and miscommunication between student and instructor.