The ideał situation is that we assembly a test version from the task bank, where the average of partiaI cut-off scores is equal to 60%.
X |
X |
x |
X |
^U1 ©— |
_U2 |
_U3 © |
i*JJ4 © |
Here, the cut-off score of the assembled test version is lower than necessary, 56%, which is why some tasks are replaced with others, ifavailable - we model the test version in order to come as close as possible to the 60% cut-off score.
The new test version has a cut-off score very close to 60%. This version is usedfor live testing.