Chris Olsen wrote: > First of all, I have no clue how one would define grading on the curve. > ... > My preferred method is to construct tests & quizzes in a way that gives > an approximately normal distribution, weight their z-scores, and sum > to a result.
Sounds as though you have a pretty good grasp of it. Before I quit trying to teach Math/Stats to unprepared/untalented studentites, and got a so-called "real" job, I used score-clustering to assign grades. My class sizes were large enough and heterogeneous enough to result in three to five clusters, and there was no doubt that the few top students [sometimes only one] earned an "A", so I never faced the task of deciding whether everyone would get a "C" or lower, but I was *much* more willing to do that than to do what so many of my colleagues did: set arbitrary grade break-points, and then subsequently fudge student performances by discarding low test scores, or by concocting low ad_hoc weightings for them. ================================================================= Instructions for joining and leaving this list and remarks about the problem of INAPPROPRIATE MESSAGES are available at http://jse.stat.ncsu.edu/ =================================================================