傳統的選擇式聯合分析法主要針對整體市場的偏好參數做推論，無法準確的針對個人層次的參數做推論。層級貝氏的方法可有效推論個人層次參數，因其模式結合整體與個人偏好的資訊，並以整體資訊作為先驗分配以輔助個人層次參數的推論。然而，我們認為若進一步以區隔的資訊作為先驗分配，應可再提升參數的準確率。基於此觀點，本研究提出設定消費者異質性偏好服從具有潛藏區隔特性的混合常態分配模型，改進過去層級貝氏方法對消費者異質性服從常態分配假設的限制。同時，並提出此模型的市場區隔方法。因此，本研究模型主要優點為可從一次的分析中，產生個人層次偏好係數、市場區隔層次偏好係數，以及市場區隔大小的資訊。最後本研究以旅遊產品為例，說明此模型在選擇式聯合分析法的應用。

Doing choice-based conjoint analysis, it is a typical approach to estimate part-worths at the aggregate level. However, hierarchical Bayes approach could overcome the limit. The Bayesian approach integrates the aggregate level parth-worths as prior information to adjust the individual level part-worths and thus individual level part-worths could be inferenced more precisely. Theoretically, the precision of individual level part-worths could be improved if the segment level part-worths are used as prior information. In view of the above reason, a hierarchical Bayes choice model with mixture of normals prior is proposed to improve the prior for the inferenc of individual part-worths. The proposed model overcomes the limits of traiditional Bayesian approaches because the traiditional approach could not provide both segment and individual levels part-worths in a conjoint analysis. In this paper, the authors illustrate how multivariate mixture of normals model could improve the understanding of those latent segments hidden in the data. Specifically, the model provides a solution for the understanding of individual preference and identifies consumer segments in an analysis. With an application to a travel service conjoint study, we show how this modeling approach could help us to uncover individual and segment levels parameters in a choice-based conjoint analysis.