Choice modeling and product design in a competitive environment
A Random partworths model is presented which takes into account similarity/dissimilarity between alternatives when determining choice probabilities for individuals. The model is estimated using generalized least squares procedure. A method is developed for optimal product selection based on these random partworths of individuals. Monte Carlo results show that there are considerable differences in choice probabilities predicted by the RP model and two commonly used alternate models, the Multinomial Logit Model, and the Maximum Utility Rule. These results also show differences in shares predicted by the various models, with greater differences being between the RP model and the MUR. Analysis of Optimal product selection is done through Monte Carlo simulation, and these results show that the recommended optimal product is often different depending on which choice model is used. The consequence of choosing the wrong model are large when the population is homogenous, or highly segmented, and when the possible alternatives are comparable to each other. Analysis of some experimental data shows some cases where the choice model doesn't make a large difference, and others where the choice model does matter.
Bansal, Pradeep Kumar, "Choice modeling and product design in a competitive environment" (1991). Dissertations available from ProQuest. AAI9211903.