Methods of Psychological Research - Online vol:6 issue:2 pages:159-174
In this paper it is shown that differences in item discrimination may be explained in some situations by means of non-modeled interactions between the items. If two items have large estimated discrimination parameters, this may be the result of a positive interaction between them. Conversely, if two items have small estimated discrimination parameters, this may be the result of a negative interaction between them. The effect is shown to be present in two data sets.