Reflections on “Stochastic Preference Theory: Some Unresolved Questions”
回应了对随机偏好理论的批评，指出独立性假设在现实中罕见，并强调后续建模已超越原始模型，适合关注偏好理论与品牌转换的学者判断是否阅读原文。
three conclusions in Preference Theory: Some Questions are entirely correct. These conclusions are: First, the theory is unable to estimate brand switching in independent subgroups. Herniter's entropy solution appears stronger in this respect. Second, stochastic preference theory provides no definitive way to estimate go, the probability of stochastic choice, for all market structures. Third, estimates of switching based on general stochastic preference do not correspond to estimates based on subgroup preference. first and the third conclusions refer to independence. I do not feel that these are serious defects because, as pointed out in the article, independence would rarely be encountered in the real world. With respect to the second conclusion, the fact that the suggested method for estimating go would not be satisfactory in all cases was mentioned in the original article. As there is no fundamental disagreement with the criticism, in what way is a reply justified? two main purposes of this reply are: (1) to point up broader and more general issues not discussed in Unresolved Questions and (2) to point out that developments subsequent to my original article have, in certain respects, moved the modeling far beyond the model originally described. My purposes in writing The Theory of Stochastic Preference and Brand Switching were: (1) to point up philosophical and practical issues concerning choice behavior and its determinants, (2) to discuss Herniter's entropy model as a basis for estimating switching, including the manner in which he structured the market, (3) to develop a simplified theory which would capture the essence of the entropy model and at the