An Analysis of Correctional Education GED Essays.

Three judges rated General Educational Development (GED) student essays that had been written by nine incarcerated youths. If judge's ratings were not consistent, students would receive a biased average rating. Traditional classical measurement theory computes an intraclass reliability coefficient to determine if judges' ratings are reliable. In contrast, a latent trait measurement theory approach can determine a student's fair average rating adjusted for any bias in the judges ratings. A comparison of these two approaches with this sample indicated that the latent trait approach yielded more useful information for training judges to be more consistent and less biased in rating GED student essays. An appendix contains the scoring guide, computer screen illustrations, and the Rasch measurement programs. (Contains 18 references.) (Author/SLD)