Requirements
Christine E. DeMars · Oxford University Press eBooks · 2010
This chapter describes the types of questions for which item response theory (IRT) is useful, along with the data requirements needed to apply IRT appropriately. It focuses on the application of IRT, not methodological research in IRT. Further, it stresses on test development and scoring. For a two-parameter logistic (2PL) model or three-parameter logistic (3PL) model with fixed c-parameters, if the θ distribution is reasonably normal and the items are highly discriminating, samples of 500 examinees and twenty items would likely be adequate for estimating both item parameters and θs. Minimum samples of at least 1,000 examinees and forty items would be a more cautious guideline for situations in which the θ distribution may not be normal or the items may not be very discriminating, or if the c-parameter is estimated rather than fixed or tightly constrained.