09 IRT for force choice items (present by Chenwei)

Wayne's comment

Wayne's comment

by CHEN Chia Wen -
Number of replies: 1
Although using forced choice type items can avoid faking responses corresponding to social expectation, there is a problem for a long time. Ipsative data causes one's score can't be comparable with other's. It means the scores can be just used at the comparison among inner-person traits . In this study, the author said if two items keyed in the same direction designed in a block, actually we just got the relative position of the two traits. However, when we mixed positive and negative keyed items in one block, this problem can be alleviated. The author explained that because the opposite direction provided more information in function 21 and 22, two traits involed can be located in absolute values. In the function 21 and 22, the parameters BETA_i and BETA_k include the factor loading parameter. When the direction between item_i and item_k factor loading is opposite, the information on sum of two traits involed are provided. Therefore, the culculating result in function 21 and 22 would be larger than the same direction. The absolute location can be gained by this way. In Table 3, it was demonstrated through +/- items had greater recovery of parameters. My question is that, It is hard to understand why does more information can produce absolute location? I mean the original data is still ipsative data type. There should be no information about inter-person comparison. Actually the opposite direction had more information in the function, but how does it let data can be compared between different individaul. I didn't find out a practical example to illustrate this effect.
In reply to CHEN Chia Wen

Re: Wayne's comment

by LIU CHEN WEI -
I think the problem is whether the definition of "negative item" is right in this paper. They assumed that the negative item has minus slope parameter. So it gives more information and help discriminating the position of latent trait (e.g., a+b=10 and a-b=2)