Evaluation Reforms: Moving From Absolute to Relative Grading with Q-Factor Control

Abstract

In the evaluation reforms, a key component is the changeover from absolute to relative grading system. In the learner-centric education paradigm, the relative grading offers certain advantages. While many premier institutes tend to follow a relative grading system, it is important to check the applicability of this shift in the context of institute-specific requirements and controllable parameters for any institute planning to change from absolute to relative grading. In this article, analyses of grade distribution and comparison of evaluation methods – the absolute grading and four types of relative grading is presented using two data sets. Marks scored by a group of 300 students in two courses where, one is a highscoring subject and the other is a low-scoring subject are considered. The four types of relative grading systems presented here are scale-up: with and without scaling factor, percentile method used in most of the competitive examinations and the normal distribution method using the median and standard deviation. All the graphs are converted to bell-curves for uniformity in comparison. The analyses shows when one grading method is suitable for a particular course, it does not yield similar result with some other course/s. While each system including the absolute grading has its own advantages, the appropriate choice of a particular system and its success depends upon the flexibility it offers. In this article, the normal distribution method is further modified to add a quality factor (Q-factor) to maintain a minimum level of competency that the students are expected to attain. With this modification, it is showed that this modified normal distribution method of relative grading can be tailor-made to diversify evaluation and grading standards. The implementation of a particular system depends upon the requirements of the institute, its attributes, effectiveness and productivity of examination system and hence these models need to be customized accordingly.