GPA算法の比較検証 : 従前のGPA からfunctional GPAへの移行とその最適互換性をめぐって

概要

論文の詳細を見る
The GPA (Grade Point Average) is becoming more common among Japanese universities as an index of academic performance. However, in many cases, raw scores are assessed based on a 100-point scale, and if the GPA is calculated by a method using letter grades that correspond to raw score ranges, the resulting GPAs represent unfair evaluation. This issue needs to be resolved for various reasons including for precise performance evaluation. Study 1 of the present research conducted 10 simulations in which grades were generated with random number generation by a computer on the assumption that a group of 1,000 took standard courses for four years. The results showed the average incidence of rank order shift of 96.9%, with the maximum shift of 97 ranks. This disagreement is resolved by taking the functional GPA method in which the GP is calculated directly from raw scores. It is desirable to adopt a computation method with the optimal compatibility, taking such issues as transition from the traditional GPA computation method. In Study 2, four representative computation methods were compared using the same simulations as Study 1: (1) GP = (raw score – X)/10, with X = 50; (2) with X = 60; (3) with X = 55; and (4) GP = (raw score – 60) x .075 + 1. The resultsdemonstrated that Method 3 was found to show the optimal compatibility with the traditional method. In many universities, distributions of grades tend to bias higher scores. Considering this tendency, to recompare Methods 3 and 4, simulations identical to Study 2 were conducted in Study 3 so that the distribution with higher scores bias. The results yielded even a lower compatibility for Method 4, while showing a sustained compatibility for Method 3, reiterating the optimality of Method 3.
2012-04-30

GPA算法の比較検証 : 従前のGPA からfunctional GPAへの移行とその最適互換性をめぐって

スポンサーリンク

概要

著者

関連論文

スポンサーリンク