质的数据的变量如何检验其中介效应？

zhouluyang

Kenny：
6 e5 o" x1 d& \/ [+ v; V       我这里本科生的毕业设计有外文翻译一事。我就让学生翻译一论文，同时模仿其研究模式进行相应的类似的研究，从而形成学士论文。
       最近学生所译的论文，我细读后，有一处百思不得其解。文中，使用质性数据的变量，并检验其中介效应的显著性。我感觉，这个无法操作呀。
       下面的摘抄相关的段落，请Kenny解答一下。
% D. w, _5 V9 n5 u7 `; T* v9 o2 a" k
4 {! w+ q0 ?  Q: q9 ^6 T: R$ y2 j+ ^# K+ F

Brand extension thoughts. We analyzed thoughts the participants expressed about the brand extension. Using definitions of analytic and holistic thinking, independent coders coded thoughts into these two categories (inter-rater reliability = 87%); disagreements were resolved by discussion. For example, thoughts were coded as analytic when they made reference to attributes (“The wallets would smell of gas and car smell”) or product category similarity (“Toyota wallets are a bad idea because cars are so different from wallets”). Thoughts were coded as holistic when they made reference to more general relationships, such as overall brand reputation (“Mercedes-Benz has luxury—they could pull off something really upscale and superior,” and “Toyota is a good brand—they would make good wallets”) or complementarity of use (“A good product. The reason is people who are held in high social class who buy Mercedes will most likely buy a Mercedes wallet,” and “People who drive need wallets too. Toyota would sell wallets to the car buyers”).1

[1The distinction between analytic and holistic thinking is not comparable to distinctions between relational versus item-specific processing or category versus piecemeal processing.]

品牌延伸联想（Brand extension thoughts）。我们分析了被试所表达的对品牌延伸的联想。依据整体性思维与解析性思维的定义，独立的编码者将这些品牌延伸联想划归为两个类别（间信度（评分者间信度）=87%）；编码结果中的分歧通过讨论得到解决。举个例子。当延伸联想指向属性（“钱包有汽油的气味和汽车的气味”）或产品类别相似性（“丰田钱包不是一个好主意，因为汽车与钱包是如此不同”）时， 这些联想就被划归为解析性思维。当延伸联想更倾向于总体性的关系时，如总的品牌声望（“奔驰豪华——他们可以成功推出一些高档的、优越的东西”，或者，“丰田是一个很好的品牌——他们能做好钱包”）或用途上的互补（“一个好的产品。这是那些在上流社会、买奔驰的人最有可能购买奔驰钱包的原因”，或者，“驾车的人也需要钱包。丰田将销售钱包给汽车购买者”），这些联想就被划归为整体性思维。

【原文注：解析性思维和整体性思维之间的区别，不同于处理关系与处理具体项目之间的区别，也不同于处理事物大类与处理单个事物之间的区别。】

Next, we examined differences in analytic and holistic thoughts. As we expected, for the functional brand, analytic and holistic thinkers exhibited differently types of thought. Analytic thoughts were greater for analytic than for holistic thinkers(80%versus35.3%), whereas holistic thoughts were greater for holistic than for analytic thinkers (64.7% versus 20%) (z=3.06, p<.01). In contrast, for the prestige brand, no differences in thoughts emerged for analytic versus holistic thinkers (analytic thoughts: 28.6 versus 11.1%; holistic thoughts: 71.4% versus 88.9%; p>.10).

接着，我们检验了解析性思维和整体性思维的不同。如我们所预期的，对于功能型品牌，解析性思维者和整体性思维者给出了不同类型的联想：解析性思维者比整体性思维者给出了更多的解析性联想（分别占80%与35.3%），而整体性思维者比解析性思维者给出了更多的整体性联想（分别占64.7%与20%）(z=3.06,p<.01)。与此形成对比的是，对于声望型品牌，解析性思维者与整体性思维者的这两类联想在数量上没有显著的不同（解析性联想分为占28.6%与11.1%；整体性联想分别占71.4%与88.9%，p>.10）

Next, we conducted a mediation analysis to examine whether extension thoughts (analytic versus holistic) mediated the effect of style of thinking on extension evaluations. Following Baron and Kenny (1986), we performed a series of regression analyses. First, we found that the parent brand concept × style of thinking interaction predicted extension evaluations (β = 1.21, t = 1.94, p = .05(可能原文有误，当为P<0.05)). Second, the parent brand concept × style of thinking interaction predicted extension thoughts (β = –.39, t = –2.07, p < .05). Finally, when we regressed the parent brand concept × style of thinking interaction and extension thoughts on extension evaluation, the effect of extension thoughts remained significant (β = –1.25, t = –3.55, p < .05), while that of the parent brand concept × style of thinking interaction dropped to nonsignificance (β = .42, t = .68, p > .10). Thus, extension thoughts perfectly mediated the effect of processing style on extension evaluation.

随后，我们进行了中介效应分析，以研究延伸联想（解析性联想与整体性联想）[extension thoughts (analytic versus holistic)]对思维模式的延伸评价效应是否起中介作用。我们按照Baron and Kenny (1986)的做法，进行了一系列的回归分析。首先，我们发现，母品牌概念×思维方式的交互项是品牌延伸评价的前因变量（β=1.21，t=1.94，p<0.05）。其次，母品牌概念×思维方式的交互项是延伸联想的前因变量（β = –.39, t = –2.07, p < .05）。最后，当我们用（母品牌概念×思维模式的交互项）与（延伸联想）这两个变量对（延伸评价）进行回归时，发现，延伸联想的影响仍然显着（β = –1.25, t = –3.55, p <0 .05），而（母品牌概念×思维方式的交互项）的影响则不显著（β =0 .42, t =0 .68, p >0 .10）。因此，延伸联想是思维模式与延伸评价之间的完全中介变量。

- L  V( k! w4 ~- X+++++++++++++++++++++++++++++
5 |+ @5 C% D8 }: [; ^+ d" {以上是原文及其翻译。
* F% X2 e; \6 V: L, h最后一段的中介效应分析，让我感觉无法做。因为extension thoughts在这里是定性数据，而（the parent brand concept × style of thinking interaction ）与extension evaluation为7点量表测量的数据。他们如何能够放在一起进行回归呢？
0 u, C' T- v$ |- f% b* g% p++++++++++++++++++++++++++
" r6 |( {* P1 k1 A% Z为了便于理解，我将原文与译文均附于此。
原文为PDF格式，可能由于我的权限问题，不让上传。不过，下面的DOC文档中也包含英文原文，只是看起来不太美观。原文为：Monga, A.B. and D.R. John, What Makes Brands Elastic? The Influence of Brand Concept and Styles of Thinking on Brand Extension Evaluation. Journal of Marketing, 2010. 74(3): p. 80.
! _2 F$ Q' i, c: E) A% ]; S2 b
1 Q% k! X$ r) O5 W. n1 Y
3 a0 C- a( q6 |6 ~) H( s

zhouluyang

感谢中人网的工作人员，亲自为我上传附件。

Kenneth

zhouluyang, 对不起，
（1 ）我读不到你的 PDF 档，只可以看你的帖子。
（2）我不是作者，自然没有能力解释为什么他们这样做。你为什么不直接问作者呢？
（3） 我也看得一头雾水。中介分析 （不是调节或交互作用）为什么会用 A*B 呢？
所以，对不起，我不知道如何回答你的问题。