Generalizing the Rasch Model to account for omitting behaviour
Lucio Bertoli Barsotti, Antonio Punzo · 2012
A two-dimensional IRT model is introduced for binary data in the presence of omitted responses. Its parameterization has the advantage to satisfy the following conditions: firstly, the tendency to omit depends on a person latent trait which is distinct from the usual latent ability; secondly, the model belongs to the Rasch family of models, guaranteeing the existence of non-trivial sufficient statistics for the parameters. An application to a real dataset is illustrated.