Insights and pitfalls - Selection bias in qualitative research
Insights and pitfalls - Selection bias in qualitative research
复制标题
DOI：
10.1353/wp.1996.0023
复制
发表时间：
1996-10-01
期刊：
影响因子：
5
通讯作者：
Mahoney, J
中科院分区：
文献类型：
--
作者：
Collier, D;Mahoney, J
定性分析人员已经收到严厉的警告，他们的研究的有效性可能会受到选择偏差的影响。本文为国际和比较研究领域的定性研究人员提供了这一问题的概述，重点关注研究者故意选择案例可能导致的选择偏差。例子来自革命、国际威慑、通货膨胀政治、国际贸易条件、经济增长和工业竞争力的研究。本文首先探讨了定量研究中关于选择偏差的见解如何最有效地应用于定性研究。讨论考虑了为什么定性研究人员需要关注选择偏差，即使他们不关心他们的发现的普遍性，它还考虑了这种形式的偏见对定性研究的独特影响，如在被标记为“基于极端情况的复杂性”的问题中。然后，文章考虑了最近关于定性研究中选择偏差的讨论中的陷阱。这些讨论有时会陷入对因变量如何概念化以及适当的比较框架应该是什么的分歧和误解中，这些问题对于评估给定研究中的偏差至关重要。在某些情况下，很明显，真正的问题不仅仅是选择偏差，而是在不同分析目标之间的更大的权衡。
Qualitative analysts have received stern warnings that the validity of their studies may be undermined by selection bias. This article provides an overview of this problem for qualitative researchers in the field of international and comparative studies, focusing on selection bias that may result from the deliberate selection of cases by the investigator. Examples are drawn from studies of revolution, international deterrence, the politics of inflation, international terms of trade, economic growth, and industrial competitiveness. The article first explores how insights about selection bias developed in quantitative research can most productively be applied in qualitative studies. The discussion considers why qualitative researchers need to be concerned about selection bias, even if they do not care about the generality of their findings, and it considers distinctive implications of this form of bias for qualitative research, as in the problem of what is labeled ''complexification based on extreme cases.'' The article then considers pitfalls in recent discussions of selection bias in qualitative studies. These discussions at times get bogged down in disagreements and misunderstandings over how the dependent variable is conceptualized and what the appropriate frame of comparison should be, issues that are crucial to the assessment of bias within a given study. At certain points it becomes clear that the real issue is not just selection bias, but a larger set of trade-offs among alternative analytic goals.