A Generalized Additive Model for Discrete-Choice Data
放松了多项Logit模型中效用函数线性参数的假设，用解释变量的一维非参数函数之和来替代，并通过模拟和消费者面板数据验证了该方法能恢复各种形状的非线性效用，对市场营销有启示。
The usual assumption of a linear-in-parameters utility function in a multinomial logit model is relaxed by a sum of one-dimensional nonparametric functions of the explanatory variables. The model generalizes the logistic regression of the generalized additive model for a binary response to a qualitative variable that can assume more than two values. Simulation studies show that the proposed method can recover underlying nonlinearity in utility of various shapes. The model is applied to consumer panel data collected by bar-code scanners from two product categories, and the marketing implications are sought.