This will result in a single percentile series for each paper, which reflects a candidate’s performance in all tests conducted under CUET-UG for that subject, through what is called the equipercentile method. This should take into account the differences in difficulty levels between the tests, according to the assessment process document, accessible only by TOI.
A candidate’s scorecard will have “normalized” percentiles and scores, not raw grades, for each subject. The universities will use the normalized grades for the preparation of the admission ranking.
Explaining the methodology, Scholarships Commission of the University president M Jagadesh Kumar said: “In the equipercentile method, we use the same scale for all candidates regardless of the session in which they appeared in a given topic, making their performance comparable between sessions.”
Over nine million graduate applicants from around 90 universities are awaiting their CUET-UG results, which the UGC he said it will be declared by September 15.
In this year’s CUET-UG, the challenge before the UGC was to prepare a single ranking for admissions, although rehearsals took place over multiple shifts on different days. CUET-UG was conducted in 27 different subjects with freedom for candidates to opt for a combination of these subjects. Scores are to be used by several universities across the country for admission to UG programs.
“How would we compare the performance of different students who wrote the test in the same subject but on different days? We had to make sure that admissions were made on the basis of a score that accurately compares student performance, ”said M Jagadesh Kumar, UGC president.
“In addition to the difficulty mentioned above, in subjects such as sport or fine arts, some universities give a certain weight (say 25%) to the skill component. However, it is not possible to perform additional raw scores of the skill component and the remaining weight age (75%) of the percentile to prepare the ranking because it would be similar to adding oranges to apples.
“One solution to this situation is the use of the equipercentile method. In this, each candidate’s normalized grades are calculated using the percentiles of each group of students in a given session over multiple days for the same subject, ”Kumar said.
To calculate normalized grades, each student’s raw grades are used to determine the candidate’s percentile on their exam round. For example, suppose 100 students showed up for the test in a given shift. Their grades are sorted in descending order. Suppose a student scored 87%. Now suppose that 80 out of 100 students have obtained grades less than or equal to 87%. This student’s percentile would be 80/100 = 0.8. The percentile will always be between “0” and “1” and is usually rounded to the required number of decimal places.
Now, suppose the single subject test ran over 10 shifts. Each student will have appeared in one shift and would be “absent” in nine rounds. The raw scores in these “absent” rounds for each student are calculated using a statistical method called “interpolation”. Once raw grades have been assigned to each student in all rounds in which the subject exam took place, these grades are averaged and on the basis of these average grades a percentile is reached. A “withdrawal” of the percentile score is also necessary for a distribution that would be close to the actual “observed distribution”. This withdrawal score is the “normalized” score.
Kumar said: “Each percentile value of candidates sorted in descending order will have raw grades for all shifts. We then calculate the average of the actual raw marks in one shift and the raw marks obtained using interpolation in other rounds. This will give the normalized grades for each candidate’s percentile This method has proven accurate for estimating normalized candidate grades when testing takes place over multiple sessions in a given subject.
