Robustifying the Rasch model with the forward search

We introduce a forward search method for identifying atypical observations in Item Response Theory models for binary data (Rasch models). Our proposal introduces diagnostic tools, based on robust high-breakdown methodologies, to avoid distortion in the estimation of the model, and to single out outlying response patterns. Atypical response patterns usually deserve further investigation. Methods to initialize, progress, and monitor the Forward Search are explored. The simulated dataset showcases the effectiveness of the method in the presence of outliers.