Abstract:
This research aimed to develop a method for adjusting cut-off scores based on
Angoff's concept by applying Item Response Theory (IRT), to compare cut-off scores based on
the traditional Angoffs concept and a new model developed by applying IRT, as classified by
experience, education level, and learning-strand, and then to develop a program for
calculating cut-off scores based on Angoff's concept by applying IRT. There were three phases
involved in the research: 1) to study and develop methods to determine cut-off scores with
IRT, 2) to compare the cut-off points, and 3) to develop a computer program for calculating
the cut-off score.
The research results revealed that:
1. The method of adjusting the cut-off score based on Angoff's concept by applying
IRT involved five components: 1) item selecting, 2) item analysis by using IRT, 3) item sorting,
4) training for determining the cut-off score, and 5) having judges consider the probability of
answering each item under three rounds of judgment.
2. The judges assessment of the cut-off score based on the traditional Angoffs
concept and the new model developed by applying IRT differed according to experience and
level of education. The judges found that, overall, the methods did not differ. However, there
were two learning strand groups: science and social studies, and religion and culture, where
judges found statistically significant cut-score differences at the .05 level by education level.
3. Comparing the cut-off score based on the traditional Angoffs concept and the
new model developed by applying IRT: there were three learning strand groups for the cut-off
scores: social studies, religion and culture, mathematics and science, in which both methods
had statistically-significant different cut-off scores at .05 levels, however no language (Thai and
foreign) differences were found.
4. The program calculated the cut-off score based on Angoffs concept by applying
the IRT, the items were appropriate at a high level and they were accepted by experts and
program users.