考慮屬性交互作用影響之離散選擇分析及其應用

Abstract

When decision makers are facing a set of choices, study of how the socioeconomic and alternative attributes affect the decisions is referred to as Discrete Choice Analysis (DCA). DCA has been developed for years and has been widely applied to various areas such as transportation planning, marketing, consumer psychology, behavior study, etc. However, conventional DCA studies rarely discuss the interaction effects among socioeconomic and/or alternative attributes on the decision making. In this research, we attempt to take the interaction effects into consideration and propose methodologies to incorporate the interaction effects into the multinomial logit models, one of the most used models in discrete choice analysis, and probe into the reasons how the interaction effects influence decision makers’ choices. We will demonstrate the proposed methodologies through two real cases, where ice cream flavors and traveling transportation, respectively, are to be chosen by consumers. The multinomial logit model with interaction effects is tested using the PHREG procedure of SAS software to show how inclusion of interaction effects leads to a very different interpretations of the choice behaviors.