Useful mathematical relationships embedded in Tversky's elimination by aspects model

It has been over 30 years since Tversky's elimination by aspects (EBA) model was introduced. Despite the potential importance of this elegant model, no method for parameter estimation has emerged. In this paper we prove that simple differences in observed choice probabilities across choice sets are linearly related to the parameters in the EBA model. Once the appropriate probability differences are calculated it is then easy to use widely available least-squares software to estimate the EBA parameters. This paper also corrects a misconception about the number of free parameters in the EBA model. The parameter estimation approach is demonstrated on data from repeated choices of one individual. (C) 2003 Elsevier Inc. All rights reserved.