grad-case-study
用 Yin 的方法設計個案研究，含三角檢證
中文翻譯
--- name: "grad-case-study" description: "運用個案研究設計（Yin），以單一或多重個案設計與三角檢證，探究真實情境中的「如何」與「為什麼」問題。當使用者需要在單一與多重個案設計之間做選擇、套用整體性與嵌入式的分析單位、從多個來源對證據做三角檢證，或問到「個案研究該怎麼架構」、「單一個案還是多個個案」或「怎麼確保個案研究的效度」時使用。" metadata: category: "WP-32 質性方法" tags: ["case-study", "Yin", "single-case", "multiple-case", "triangulation", "holistic", "embedded", "replication-logic"] --- # 個案研究（Yin） ## 概述 個案研究是一種經驗性的探究，在現象的真實情境中深入調查一個當代的現象，尤其適用於現象與情境之間的界線並不明顯的時候。Yin 的框架提供了系統化的設計選擇（單一 vs. 多重個案、整體性 vs. 嵌入式分析），並強調用三角檢證來強化建構效度。 ## 何時使用 - 研究問題是關於當代事件的「如何」或「為什麼」問題 - 研究者對行為事件幾乎沒有或完全沒有控制 - 情境條件與研究的現象高度相關 - 現象與情境之間的界線很模糊 ## 何時不要用 - 需要頻率或發生率的資料時（用問卷調查或實驗） - 情境無關緊要、而且可以被控制時（用實驗） - 目標是對母體做統計上的推論時 ## 假設 ``` IRON LAW: Case study answers HOW and WHY questions in context — if you need frequency or incidence data, use a survey or experiment instead. Applying case study to "how many" or "how much" questions misuses the methodology. ``` 關鍵假設： 1. 個案是一個有邊界的系統，要定義時間、空間和概念上的邊界 2. 多個證據來源對建構效度是不可或缺的 3. 多重個案遵循複製邏輯（不是抽樣邏輯），每個個案是一次實驗，不是一位問卷填答者 4. 個案研究的草案（protocol）和資料庫能提升信度 ## 方法論 ### 步驟 1：設計個案研究 定義研究問題（如何/為什麼）。用 Yin 的 2x2 矩陣選擇個案類型： | | 單一個案 | 多重個案 | |---|---|---| | **整體性**（單一單位） | 關鍵、獨特或揭露性的個案 | 逐字（literal）或理論性的複製 | | **嵌入式**（多個單位） | 一個個案內的多個單位 | 跨個案的多個單位 | 發展命題來引導資料蒐集。 ### 步驟 2：從多個來源蒐集證據 從六種來源類型中，至少蒐集三種的資料：文件、檔案紀錄、訪談、直接觀察、參與觀察、實體人造物。維持一條證據鏈，把問題連結到資料再連結到結論。 ### 步驟 3：運用三角檢證 | 三角檢證類型 | 說明 | |---|---| | **資料** | 多個資料來源收斂到同一個發現 | | **研究者** | 多位研究者獨立分析同一份資料 | | **理論** | 對同一份資料套用多種理論視角 | | **方法論** | 多種方法（質化 + 量化）處理同一個問題 | ### 步驟 4：分析與報告 使用型態比對、解釋建構、時間序列分析或跨個案綜合。要透明地呈現證據鏈。 ## 輸出格式 ```markdown ## Case Study Analysis: [Context] ### Research Question - Question: [the how/why question] - Case type: [single/multiple] x [holistic/embedded] - Unit of analysis: [what constitutes the "case"] ### Case Selection Rationale | Case | Rationale | Expected Pattern | |------|-----------|-----------------| | [name] | [why selected] | [literal/theoretical replication] | ### Evidence Matrix | Source Type | Data Collected | Key Findings | |------------|---------------|--------------| | [documents/interviews/etc.] | [description] | [finding] | ### Triangulation Results - Convergent findings: [where sources agree] - Divergent findings: [where sources disagree] - Explanation: [how divergence is resolved] ### Pattern Matching - Predicted pattern: [from propositions] - Observed pattern: [from evidence] - Match assessment: [strong/moderate/weak] ### Conclusions 1. [Key finding with chain of evidence] 2. [Analytical generalization — how findings extend theory] ``` ## 常見陷阱 - 個案研究推論的對象是「理論」（分析性推論），不是母體（統計性推論） - 單一個案設計需要明確的理由：關鍵、極端、獨特、揭露性或縱貫性 - 多重個案中的複製邏輯，意味著每個個案都必須獨立地證實或否證一個命題 - 證據鏈必須能從問題追溯到草案、資料庫，再到報告 - 「不要」把個案研究跟病歷（case history）或個案報告（case report）混淆，Yin 的個案研究是有正式設計的研究策略 - 競爭的解釋必須明確處理，不能只是輕輕帶過 ## 參考資料 - Yin, R. K. (2018). *Case Study Research and Applications: Design and Methods* (6th ed.). Sage. - Stake, R. E. (1995). *The Art of Case Study Research*. Sage. - Eisenhardt, K. M. (1989). Building theories from case study research. *Academy of Management Review*, 14(4), 532-550.
英文原文
--- name: "grad-case-study" description: "Apply case study research design (Yin) to investigate how and why questions within real-life contexts using single or multiple case designs and triangulation. Use this skill when the user needs to choose between single and multiple case designs, apply holistic versus embedded units of analysis, triangulate evidence from multiple sources, or when they ask 'how should I structure a case study', 'single or multiple cases', or 'how do I ensure validity in case research'." metadata: category: "WP-32 質性方法" tags: ["case-study", "Yin", "single-case", "multiple-case", "triangulation", "holistic", "embedded", "replication-logic"] --- # Case Study Research (Yin) ## Overview Case study research is an empirical inquiry that investigates a contemporary phenomenon in depth and within its real-life context, especially when the boundaries between phenomenon and context are not clearly evident. Yin's framework provides systematic design choices — single vs. multiple cases, holistic vs. embedded analysis — and emphasizes triangulation to strengthen construct validity. ## When to Use - Research questions are "how" or "why" questions about contemporary events - The researcher has little or no control over behavioral events - Contextual conditions are highly pertinent to the phenomenon of study - Boundaries between phenomenon and context are blurred ## When NOT to Use - When frequency or incidence data is needed (use survey or experiment) - When context is irrelevant and can be controlled (use experiment) - When the goal is statistical generalization to a population ## Assumptions ``` IRON LAW: Case study answers HOW and WHY questions in context — if you need frequency or incidence data, use a survey or experiment instead. Applying case study to "how many" or "how much" questions misuses the methodology. ``` Key assumptions: 1. The case is a bounded system — define temporal, spatial, and conceptual boundaries 2. Multiple sources of evidence are essential for construct validity 3. Multiple cases follow replication logic (not sampling logic) — each case is an experiment, not a survey respondent 4. A case study protocol and database enhance reliability ## Methodology ### Step 1: Design the Case Study Define the research question (how/why). Select the case type using Yin's 2x2 matrix: | | Single Case | Multiple Case | |---|---|---| | **Holistic** (single unit) | Critical, unique, or revelatory case | Literal or theoretical replication | | **Embedded** (multiple units) | Multiple units within one case | Multiple units across cases | Develop propositions to guide data collection. ### Step 2: Collect Evidence from Multiple Sources Gather data from at least three of six source types: documents, archival records, interviews, direct observation, participant observation, physical artifacts. Maintain a chain of evidence linking questions to data to conclusions. ### Step 3: Apply Triangulation | Triangulation Type | Description | |---|---| | **Data** | Multiple data sources converge on the same finding | | **Investigator** | Multiple researchers independently analyze the same data | | **Theory** | Multiple theoretical perspectives applied to the same data | | **Methodological** | Multiple methods (qual + quant) address the same question | ### Step 4: Analyze and Report Use pattern matching, explanation building, time-series analysis, or cross-case synthesis. Report the chain of evidence transparently. ## Output Format ```markdown ## Case Study Analysis: [Context] ### Research Question - Question: [the how/why question] - Case type: [single/multiple] x [holistic/embedded] - Unit of analysis: [what constitutes the "case"] ### Case Selection Rationale | Case | Rationale | Expected Pattern | |------|-----------|-----------------| | [name] | [why selected] | [literal/theoretical replication] | ### Evidence Matrix | Source Type | Data Collected | Key Findings | |------------|---------------|--------------| | [documents/interviews/etc.] | [description] | [finding] | ### Triangulation Results - Convergent findings: [where sources agree] - Divergent findings: [where sources disagree] - Explanation: [how divergence is resolved] ### Pattern Matching - Predicted pattern: [from propositions] - Observed pattern: [from evidence] - Match assessment: [strong/moderate/weak] ### Conclusions 1. [Key finding with chain of evidence] 2. [Analytical generalization — how findings extend theory] ``` ## Gotchas - Case studies generalize to THEORY (analytical generalization), not to populations (statistical generalization) - A single-case design requires explicit justification: critical, extreme, unique, revelatory, or longitudinal - Replication logic in multiple cases means each case must independently confirm or disconfirm a proposition - The chain of evidence must be traceable from question to protocol to database to report - Do NOT confuse case study with case history or case report — Yin's case study is a research strategy with formal design - Rival explanations must be addressed explicitly, not just dismissed ## References - Yin, R. K. (2018). *Case Study Research and Applications: Design and Methods* (6th ed.). Sage. - Stake, R. E. (1995). *The Art of Case Study Research*. Sage. - Eisenhardt, K. M. (1989). Building theories from case study research. *Academy of Management Review*, 14(4), 532-550.