1. The item information (Stark, 2005) is related to the distance between two statements, if the participant evaluates the stimulus independently, then select the statements in the same dimension with large distance will also be benefit for person estimation. However, Table 4 showed that there was no effect for the number of unidimensional pairings and recommended to minimize their uses. Actually, the increase of total item number was relatively larger than the number of unidimensional pairings. For example, 3-D, 10 items per dimension, when percentage is 5, the uni-dimensional pairing is only 1.5 items; when increased to 20 items per dimension, the total number of items was 60, but the unidimensional items was only increased to 3 items. I think it was unfair to compare the effects of unidimensional pairings and item numbers in such condition.
2. If the locations of statements from different dimensions are comparable, then the person locations in different dimensions are also comparable, in other words, whether we can say they are actually located on the same scale? And what is the meaning by “multi-dimension”? How to interpret the item location?
3. In study 3, the correlation between theta(s) was set to be 3 levels (.0, .3, .5), and the results showed the scoring algorithm was robust. Or can it be concluded that, the independence was not a strong assumption for the model? What would happen if the correlation between theta(s) was set to be 1?
4. If one dimension always received negative response, then how to determine the person’s location on this dimension?