I am struggling to figure out how to extend DIF analysis at item level to the item group (item family or item cluster) level using Mplus. 1. With data in wide format I cannot indicate which items (columns) belong to which item group, so MIMIC with modindices seems inapplicable. 2. With data in long format, and these options I can set up a random 1PL model: USEVARIABLES = person item y itemtype; CLUSTER = person item; CATEGORICAL = y itemtype; TYPE = CROSS;
%WITHIN% %BETWEEN person% y; %BETWEEN item% y; But as multigroup is not possible with crossclassified, I wouldn't know how to compare item groups. If these types of analyses are indeed not possible, do you have any plans on implementing it within a year's time?