Monday, 21 March 2005

What we have here is a failure to discriminate

The second exam in American government had good and bad points; the good point was that the average, an 82, was basically where I wanted it to be. The bad point is that, somehow, the standard deviation was 4. So bad, in fact, I was sorely tempted to rescale the scores so there would be a larger standard deviation. Even worse, the correlation between scores on the first two exams was something like 0.2.

In layman’s terms, somehow I managed to ask 30 multiple choice questions, about 25 of which were either too easy (almost everyone got them right) or too hard (the only way people got them right was by guessing); coupled with the essay questions that never seem to discriminate well among students, I produced an exam that was borderline useless. Ugh.