关键词: 自适应设计, 经济试验, 随机化, 社交网络

Abstract: With the development of the network era, the inter-individual relationship has increasingly caught researchers’ attention. The impact of social networks has been considered in economic experiments. The existing randomization methods face two major challenges. First, participants in the experiment fail to fulfill the stable unit treatment value assumption, leading to estimation bias if the network interference is ignored. Second, more participants are expected to be assigned to the superior group from the economic consideration. In this paper, we propose a social network adjusted response-adaptive randomization to address the two issues mentioned above. The proposed method combines the response-adaptive randomization with the balance of network interference to assign more nodes to superior groups while ensuring an accurate estimation of the effect of the intervention. Both the simulation and real data illustrate that the comprehensive superiority for the proposed method, which proves a necessary tool in economic experiments in the network era.

Key words: Adaptive Design, Economic Experiment, Randomization, Social Network