对A、B两种产品进行偏好选择，怎么判断偏好出于产品特性而非实验操作导致

引言

行为实验中，给参与者呈现两根香烟（A 和 B），询问更喜欢哪种？结果会发现40%更喜欢A，20%没有偏好，40%更喜欢B。但实际上AB是相同的。那么这种偏好结果很可能是由于实验操作或其他变量引起的，而非参与者对产品的特性做出的反应。

One issue to be addressed is a tendency to report preferences when the stimuli are putatively identical. Ennis and Collins (1980) mailed two cigarettes (call them “A” and “B”) to a large number of con- sumers’ homes for comparison on a variety of attributes like: better fla- vor, easier draw, better aftertaste, slower burning, and so forth. Finally they were asked for their preferences and 40% reported preference for cigarette “A,” 20% reported “No Preference” and 40% preference for “B.” Yet, “A” had been taken from the initial part and “B” had been taken from the final part of the same production run. They were essentially the same cigarette. Therefore, any preferences expressed for one or the other of the cigarettes would have been due to factors other than their sensory characteristics. This experiment was repeated with four different brands of cigarettes with consumer sample sizes ranging 412–488 (total 1,787). There was remarkable agreement between each brand.

方法

基于信号检测论的一种验证方法。

实验设置：

1、前半部分，可比较组：对两种薯片试吃，并选出喜欢的一种。如烧烤味为p1，红烩味为p2，即<p1,p2>。两种薯片在视觉上也可以分辨出来。
2、后半部分，安慰剂组：在完成比较组的选择之后，如果参与者选择偏好p1，则安慰剂组呈现的两种薯片均为p1，即<p1,p1>。同样询问参与者更喜欢哪种薯片。（理论模拟结果如TABLE 2）

验证的简单逻辑

’Prefer p1‘为信号检测论中，信号+噪音的分布；<p1，p1>为单纯的噪音分布。
如果’Prefer p1‘和<p1，p1>的分布差异显著，则说明和p2相比，参与者对p1薯片的喜欢是真的存在偏好（受产品特性影响）。

参考文献：

[1] Zhang X, Halim J, Wichchukit S等. Paired Preference Tests: A signal detection based analysis with separate d′ values for segmentation[J]. Journal of Sensory Studies, 2016, 31(6): 481–491. DOI:10.1111/joss.12238.