Yin Case Study Research

殷《案例研究：设计与方法》

Use this skill to build a traceable argument from a bounded contemporary case to an analytic conclusion—not to treat “case study” as a synonym for anecdote, one participant, or any qualitative project.

Route the request

• 是否适用案例研究、案例定义及与其他方法区别：chapters/01-method-choice.md
• 研究设计、四种类型、质量检验与分析性归纳：chapters/02-design.md
• 研究者技能、方案、培训与试点：chapters/03-preparation.md
• 六类证据、三项原则、资料库与证据链：chapters/04-evidence.md
• 总体策略与五种分析技术：chapters/05-analysis.md
• 报告结构、读者、审阅、匿名与证据呈现：chapters/06-reporting.md
• 术语、设计模板与速查：glossary.md、patterns.md、cheatsheet.md

Core architecture

questions → propositions → case/unit boundary → evidence logic → analytic criteria → findings → analytic generalization

A design links the data to be collected—and the conclusions to be drawn—to the initial questions. Build this logic before fieldwork, then maintain justified flexibility when evidence changes the understanding of the case.

Working procedure

1. Confirm that the focal question is explanatory or descriptive, contemporary, contextual, and not dependent on behavioral control.
2. Define the case with time, place, organization, population, process, and inclusion/exclusion boundaries.
3. State propositions or purpose, unit(s) of analysis, data-to-proposition logic, and criteria for interpreting findings.
4. Choose single/multiple and holistic/embedded design; justify the choice and anticipate design drift.
5. Specify theoretical replication, rival explanations, and quality tactics before data collection.
6. Develop and pilot a protocol; train researchers to ask, listen, adapt, and preserve evidence.
7. Collect converging documents, archives, interviews, direct/participant observation, and physical artifacts as relevant.
8. Maintain a separate case database and a bidirectional chain of evidence.
9. Analyze with an explicit general strategy and appropriate techniques.
10. Tailor the report to its audience and expose enough evidence and method for review.

Decision rules

• Use experiments when control of behavior is central and feasible; surveys for prevalence or population estimates; archival/history methods for past events lacking direct actors; case study for contextualized contemporary how/why inquiry.
• The case and unit of analysis must be explicitly bounded before claiming evidence sufficiency.
• Multiple cases are replications, not a sampling frame; select them for predicted similar or contrasting results.
• Single-case designs require a strong rationale such as critical, unusual, common, revelatory, or longitudinal case.
• Embedded units enrich analysis but must be reconnected to the whole case.
• Generalize analytically to theory, not statistically from a small case sample to a population.
• Triangulation means convergence on a fact or finding, not merely using many methods.
• Always examine plausible rival explanations and present disconfirming evidence.

Preferred output

Return: question and method fit → bounded case/unit → propositions and rivals → design type/replication logic → protocol/evidence sources → quality tactics → analysis technique → report form → limits of analytic generalization.