Item Response Theory Models for Carry-Over Effect Across Different Scales.

摘要: It is common in educational and psychological tests or social surveys that the same statement judged on multiple scales. These responses are linked by statement, which may cause local dependence. Considering way a scales, new class of item response theory (IRT) models developed to account for nonrecursive carry-over effect, can be affected only its preceding rather than subsequent response. The parameters estimated with freeware WinBUGS. Two simulation studies were conducted evaluate parameter recovery consequences model misspecification. Results showed recovered fairly well; fitting unnecessarily complicated data did not have effect little harm estimation; ignoring standard IRT yielded biased estimates parameters, correlation between latent traits, test reliability. empirical examples parallel design sequential provided demonstrate implications applications models.