Events and Meetings of Italian Statistical Society, Advances in Latent Variables - Methods, Models and Applications

Font Size: 
Assessment of item correlation in Rasch models
Svend Kreiner, Karl Bang Christensen

Last modified: 2013-06-14


Item analysis by IRT and Rasch models and confirmatory factor analysis (CFA) are often regarded as very different psychometric methods. The purpose of this paper is to show 1) that there are but a few technical differences between the models underlying the analyses and that the differences has to do with the different approaches to statistical inference and in particular the tests of data-model fit, and 2) that it is easy to adopt the CFA approach during item analysis by Rasch models. The methods will be illustrated by analysis of data on reading ability in Italy collected by PISA in 2006.

Full Text: PDF