Randomized Item Response Models

Abstract

Item responses can be masked before they are observed via a randomized response mechanism. This technique is used to protect individuals and improve their willingness to answer truthfully. Various traditional randomized response sampling techniques are discussed and extended to a multivariate setting. So-called randomized item response models will be introduced for analyzing multivariate randomized response data. This class of models can also be extended to handle explanatory information at di_erent hierarchical levels. The models discussed are particularly suitable for analyzing sensitive individual characteristics and their relationships to background variables.