Abstract
This chapter describes the overall model fit. Overall fit statistics assess the adequacy of the model for a data set as a whole. Overall tests of fit are also useful for Rasch models, even though their lack of transparency concerning why the model does not fit means that such fit statistics can never stand alone. This chapter describes a conditional likelihood ratio (CLR) test derived by Erling B. Andersen in the early 1970s and illustrates its use on the data on the disinhibited eating (DE) subscale of the Diabetes Health Profile (DHP). Andersen's test represented a major step forward in the theory of Rasch models at the time. Since then, other overall tests have been proposed but Andersen's test is still unsurpassed by the competitors. The CLR test compares conditional maximum likelihood (CML) estimates of item parameters in different subgroups to the CML estimates for the complete sample of persons.
| Original language | Undefined/Unknown |
|---|---|
| Title of host publication | In:Rasch Models in Health |
| Editors | Mesbah M Christensen KB Kreiner S |
| Number of pages | 5 |
| Publisher | Wiley |
| Publication date | 4 Mar 2013 |
| Pages | 105-109 |
| Publication status | Published - 4 Mar 2013 |
