Abstract
Local response dependency has been a challenge to Rasch models (and other IRT models) from the days of their conception. The theory of Rasch models contains two relatively simple test statistics addressing the hypothesis of local independence that are statistically sound. The first test is the conditional likelihood ratio (CLR) test proposed by Henk Kelderman in his seminal paper on log-linear Rasch models. The second test is a simple non-parametric test that does not require specialized Rasch software. This test was first proposed by Kreiner and later was elaborated by Kreiner and Christensen. This chapter discusses how these tests work and how the simple approach proposed by Kreiner and Christensen addresses the issue of local response dependence in a way equivalent to Kelderman's approach.
Original language | Undefined/Unknown |
---|---|
Title of host publication | In:Rasch Models in Health |
Editors | Mesbah M Christensen KB Kreiner S |
Number of pages | 6 |
Publisher | Wiley |
Publication date | 4 Mar 2013 |
Pages | 131-136 |
Publication status | Published - 4 Mar 2013 |