How good are my grades?

Abstract

Currently no index exists for quantifying the grading effectiveness of tests or evaluation systems. The existing literature in educational measurement is suggestive of solutions; however, many of the techniques are only applicable in special cases and the theoretical and computational foundations are difficult. An alternative, the Qa statistic, is proposed. This statistic is the estimated proportion of correctly graded students. It is an extension of the standard error of measurement and takes into account student observed scores, reliability of available score data, and instructor-determined grade cutoffs. A one-page, computer-generated Qa report is analyzed and implications for its use are discussed. Data are presented from a sample of 60 grading situations showing that increased testing produces small marginal increases in proportion of correctly graded students.