12-30-2016 01:47 PM
Can anyone tell me how to take into account the mean group difference when using Proc IRT when using SAS 9.4? I am trying to do a DIF analysis using anchor items. When I use the syntax below, I get acceptable results as long as both of my groups have the same mean. However, when the group means are different I find a difference in the item parameters for non-anchor items when there should be no difference.
proc irt data=new out=irt_out SCOREMETHOD=MAP;
model R1-R40 /resfunc=Rasch;
This table shows some of my difficulties. These data are based on a simulated 40 item test, generated using the Rasch model, and both groups had identical item difficulty parameters. The mean between-group item difficulty difference for non-anchor items should be 0. I get that result, or close to it, when the group means are equal, but not when the means are different.
Does anyone know if Proc IRT assumes that the groups have the same mean ability level? Is there any way to account for the difference using Proc IRT?
Thanks in advance.