Multi-Attribute Choice Model: An Application of the Generalized Nested Logit Model at the Stock-Keeping Unit Level

Abstract

AbstractThis paper proposes an application of the generalized nested logit (GNL) model which is used in transportation science for product choice problems at the stock-keeping unit level. I explain two alternative nesting rules: attribute separation and latent-class separation based on taste heterogeneity. First, using the former nesting rule, I demonstrate that the GNL model is superior to the multinominal logit and the nested logit models in terms of reproducibility of choice probabilities. Second, using latter nesting rule, I reveal that the compromise effect, which is inconsistent with utility maximization, occurs in the GNL model, which belongs to the general extreme value family. This shows that the compromise effect is, in fact, consistent with utility maximization in random utility circumstances.KeywordsUtility MaximizationMode ChoiceGeneral Extreme ValueRoute ChoiceTransportation Research RecordThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.