Elimination by Dimensions
针对PRETREE模型在营销实践中的不足，提出维度消除（EBD）模型，识别消费者用于层级消除备选方案的关键连续维度，并在两个数据集上相比条件概率方法显著提升市场份额预测精度。
The well-known PRETREE model, though rich in behavioral structural theory, is virtually unused by marketing practitioners. Four problem areas relating to the application of PRETREE are identified. The authors then develop the “elimination-by-dimensions” (EBD) model to address these problems. A key feature of the EBD model is the recognition that most real-life marketing attributes are continuous dimensions, not aspects. The EBD model provides diagnostic information by labeling the specific dimensions consumers use to eliminate alternatives hierarchically. A procedure for naming the key dimensions directly from the data is suggested. Empirical information from both an industrial buying dataset and a consumer product dataset is used to compare the predicted market share estimates of the EBD model with predictions based on Tversky and Sattath's suggested conditional probability approach and the multinomial logit model. The EBD model gives substantially better market share predictions on both datasets than the conditional probability approach. Finally, using only dimensions identified by the disaggregate EBD model, the authors show individual-level predictions to be significantly above the chance level.