20130825
1) In Table 2, is it possible that three variables (age, gender, motivation) all achieve significant level in Node 1 respective? In the example mentioned by author, gender is significant in Node 1 rather than age and motivation. Therefore the splitting variable in Node 1 is decided as gender. However, if the statistic for gender, age, and motivation are all significant in Node 1, how to draw the Rasch tree? how to put the splitting variable in Node 1?
2) In Table 5, I thought that the cut point was specified at 50 for LR test, but how was the average estimated Cut point for LR test calculated? There should be no procedure in LR test to estimate the location of cut point.
3)why the minimal node-size is set at 20 in stopping rule? Isn't this size too small to estimate the parameter in Rasch model?
4) This model didn't detect the DIF bias in individual item. It detect whether there is any item has differential functioning. (Maybe the title "DIF" should be replaced to "DTF") Therefore, I think the Bonfferoni adjustment is still necessary for Rasch tree model even the author mentioned it don't need to use Bonfferoni adjustment in page 13. The result in Table 8 also revealed that Type I error was inflated when there is no Bonfferoni adjustment.