Disaggregate Tree-Structured Modeling of Consumer Choice Data
提出一种新方法，通过分类算法在个体层面估计决策树，无需事先假设树形结构，并汇总样本中决策规则的多样性及其对市场份额的影响，适用于面板数据分析。
A new approach to inferring hierarchical models of consumer choice is described. A classification algorithm is used to estimate decision trees at an individual level without requiring prior assumptions about tree form. Derived models are analyzed within a modeling system that summarizes the diversity of decision rules in a sample as well as their implications for aggregate market shares. An application to the analysis of panel data and a comparison with disaggregate logit analysis are reported.