DETECT is modified for test designs with unestimable item pairs such as multistage testing.
The basic idea is that if two items are in the same cluster, then the item pair has a positive contribution
to the DETECT value, otherwise it has a negative contribution.
In the simulation, only one correlation（0.8） is considered for multidimensional cases.
Do CAT have the need for dimension checking since the dimension should be known in the item banking? And it seems no stage1 for CAT.