Determining Item Parameters for Developing Better Testing Systems Abstract:
Modern day university entrance examinations witness large scale participations, all competing for comparatively very low number of seats. Such a tough scenario not only enforces a lot of pressure to the students but also sets challenges for the paper setters. The aim is to come up with a set of questions (items) which evaluated eventually would best serve the purpose of examination to extract qualitative information from the scores. In this work, we aim at predicting beforehand item parameters of toughness and discrimination using measures on cognitive faculties that a student uses to answer a question. We would further use the predictions
to help reduce the flaws contained in modern day large scale offline testing systems.
Continuing our previous work of better student evaluation by predicting student’s answer, we make an attempt over improving student evaluation by targeting present day testing systems. There are numerous large scale offline testing systems for admissions in undergraduate and graduate studies, most of which are objective based competitives. Hundreds of thousand students compete for a place among few top thousand students. Restricting the time available for testing, there can only be limited questions put forth to
the mass. This leads to a situation of a very tough neck to neck competition amongst the students. IIT-JEE conducted by Indian Institutes of Technology is the most well known examination exposing this scenario. The results are such that even at the very top ranks (1000-2000), there are about a couple hundred of students at the same score.
Situation gets even worse as we go to bottom ranks where one could drop or add thousands of ranks with a difference of about unit marks. As the purpose of the examination is to select the few best in the large mass of students,it becomes important for the paper to contain a set of questions which extracts the best students and also explores the discrimination between them minimizing the variation in the number of positions per unit difference of marks.
Download Determining Item Parameters for Developing Better Testing Systems Project Report