New Flexible Models and Design Construction Algorithms for Mixtures and Binary Dependent Variables

This thesis discusses new mixture(-amount) models, choice models and the optimal design of experiments. Two chapters of the thesis relate to the so-called mixture, which is a product or service whose ingredients’ proportions sum to one.
The thesis begins by introducing mixture models in the choice context and develops new optimal design construction algorithms for choice experiments involving mixtures. Building further, varying the total amount of a mixture, and not only its ingredient proportions, might also affect the response.

The models that exist for mixture-amount data date back to the 1980s and have several drawbacks, which limit their usefulness for these data. Therefore, the next chapter in this thesis develops new flexible models for mixture-amount data, which are based on so-called Gaussian processes. The last chapter builds on the aforementioned model and, using revealed preference data on green vehicle purchases in France, presents a new choice model that accounts for latent environmental consciousness, where environmental consciousness is allowed to have a flexible heterogeneous impact on the vehicle choice across the population.