Warrantable and Unwarranted Methods : The Case of QCA
Warrantable and Unwarranted Methods : The Case of QCA
复制标题
有保证和无保证的方法：QCA 案例
DOI：
--
复制
发表时间：
2013
期刊：
影响因子：
--
通讯作者：
Jason Seawright
中科院分区：
文献类型：
--
作者：
Jason Seawright
社会科学的方法论多元化版本需要规范来约束在关于因果推理的辩论中可能被视为可接受的一套方法。本文提出了一个可授权性标准，其实质是，只有在至少存在某种类型的研究设计情况下，才能认为方法是可用的，在这些情况下，可以很好地论证所讨论的方法是有效的。在阐述了这一标准的细微差别之后，本文通过构建一个重要的方法家族(CRISP-SET QCA)目前不能保证的案例来应用该标准。这种情况是因为cs-qca缺乏合理的分析结果，在模拟中表现相对较差，而且经常不能复制已知的因果结果。相比之下，其他方法，包括在实验中使用回归和CART来分析布尔型因果复杂性，更接近于满足可保证标准。
A methodologically pluralistic version of the social sciences requires norms to bound the set of methods that may be seen as admissible in debates about causal inference. This essay proposes a standard of warrantability, which is essentially the idea that methods should only be seen as usable if there is at least some class of research design situations in which a good argument can be made that the method in question works. After developing the nuances of this standard, the essay applies it by building a case that one important family of methods (crisp-set QCA) is currently unwarrantable. This case is made by showing that cs-QCA lacks justifying analytic results, performs relatively poorly in simulations, and frequently cannot replicate known causal results. By contrast, other methods including the use of regression in experiments and CART to analyze Boolean-type causal complexity, come closer to meeting standards of warrantability.