Pauci sed boni: An Item Response Theory Approach for Shortening Tests

Item Response Theory (IRT) is the theoretical framework often used for shortening tests. This contribution presents a new IRT-based item selection procedure which is meant for this purpose. This procedure is based on the information that each item provides in respect to different trait levels of interest (denoted as θ targets), which are obtained by segmenting the latent trait in either equal or unequal intervals. In a simulation study, the performance of the new procedure was compared with that of the typical IRT procedure and of a random selection of the items. The new procedure outperformed the other two in recovering central and peripheral regions of the latent trait continuum, particularly when the short test forms consisted of fewer items. Despite this study highlighted the potentiality of the new item selection procedure for developing short test forms, work is still needed.