Item fit, item stability, coverage of item content and person measure (item-person map), and item targeting were evaluated.
When a simple IRT model does not fit the data, one could consider deleting or changing bad fitting items, and/or deleting bad fitting persons.
In relation to fit, items with an infit mean square close to 1 are optimal, although some variation is necessary [ 20].
Look for items that don't fit, items that have gone out of style, items with stains or other signs of wear, and items that you just don't wear very often.
Small items can be choking hazards, so bring an empty toilet paper roll to fit items through if it fits without touching the sides it is a choking hazard.
Although the resultant Self-Care scale containing 16 items showed slightly elevated overall item fit residual statistics (Item Fit Resid. SD =1.87), there was no individual item misfit and no misfitting persons.
Through systematic investigation of response category ordering, model fit, item bias, and local response dependency, we succeeded in constructing a 20-item unidimensional scale of activity and participation (DM1-Activ).
The results of the overall fit, item and fit statistics referring to construct validity are shown in Table 3.
Each set of items was assessed for threshold ordering, overall model fit, item fit, person fit, reliability, differential item functioning, targeting and dimensionality.
Twelve HFS-B items showed excellent fit (item responses were as expected), with proper ranges of infit values (0.64 1.28) and outfit values (0.67 1.09).
