Application of Unidimensional Item Response Theory Models to Multidimensional Data

摘要

A simulation model was developed for generatingitem responses from a multidimensional latent traitspace The model permits the prepotency of a generallatent trait underlying responses to all simulated itemsto be varied systematically. Five levels of prepotencywere used to generate data sets The levels of prepotencyranged from a truly unidimensional latent traitspace to a very weak general latent trait. Simulateditem pools with guessing and without guessing wereanalyzed by the LOGIST computer program The generallatent trait was recovered in data sets where theprepotency of the general latent trait was only moderate.Consequently, it appears that item response theorymodels can be applied to moderately heterogenousitem pools under the conditions simulated here.