A
test score does not represent the exact ability of an examinee.
It only shows just one aspect of the examinee, even if the coverage
of the test is restricted. Due to this, for example, we may not
see obvious relationships between entrance examination scores and
academic records in universities, even in mathematics subjects.
Thus, in order to make clear such a relationship in a statistical
sense, we have investigated three testing records of the placement
test, the learning check test, and term examinations. Then, we
have shown mainly three consequences from the investigation:
1) by using the full computer based testing results of the placement test, we
have become aware of the magnitude of irreducible probabilistic fluctuations;
2) in using the description type testing, it would be inevitable to accept biased
evaluations by teachers;
3) by adopting full computer based testing in the placement test, the learning
check test, and term examinations, we can remove the teacher's evaluation bias
occurred in the description type testing, and can obtain the more accurate student's
ability.
In addition, we have proposed a fundamental ability equation on student's ability
including irreducible probabilistic fluctuations. |
|
|
|
|
ability
estimation, item response theory, computer based testing,
irreducible probabilistic fluctuation, evaluation bias, description
type testing, multiple choice type testing, academic growth,
ability equation, learning analytics. |
|
|