研究问题质量与研究设计审计
检查研究问题能否被设计与数据支持,形成测量抽样计划、偏差清单和推断边界。
完整提示词
Main objective: > Perform a rigorous Research Question Quality Audit workflow connecting question, claim, design, measurement, sampling, analysis and inference limits. ## 1. RESEARCH CONTEXT Establish domain, research aim, unit of analysis, target population/system, time horizon, data availability, ethical constraints, resources, decision stakes and intended claim type. ## 2. SPECIALIZED WORKFLOW - state the empirical question in falsifiable terms - identify population/system, exposure/intervention, comparator and outcome where applicable - separate descriptive, predictive and causal aims - define scope and boundary conditions - identify decision relevance - state what result would count against the hypothesis ## 3. CLAIM-DESIGN GATE Every major claim must map to a design capable of supporting it. Distinguish: - description - association - prediction - intervention effect - causal effect - mechanism - generalization Do not let wording outrun identification. ## 4. PRE-SPECIFICATION Capture: ```text Primary question: Primary outcome / estimand: Key exposure/intervention: Comparator: Population: Sampling: Design: Primary analysis: Exclusions: Missing-data strategy: Sensitivity analyses: Confirmatory vs exploratory: Deviation log: ``` ## 5. BIAS & VALIDITY Audit selection bias, measurement error, confounding, attrition, missingness, researcher degrees of freedom, temporal ambiguity, multiplicity and external-validity limits. ## 6. REPORTING STANDARDS Use design-appropriate reporting guidance. For example, CONSORT 2025 applies to randomized trial reporting, STROBE to major observational designs, and PRISMA 2020 to systematic reviews. Reporting checklists improve transparency but are not themselves proof of methodological quality. ## 7. REQUIRED MATRICES ### Question-Design Matrix | Claim | Design | Required data | Main bias | Identifiable? | Limitation | |---|---|---|---|---|---| ### Variable Matrix | Construct | Operational measure | Role | Timing | Validity | Error risk | |---|---|---|---|---|---| ### Assumption Register | Assumption | Why needed | Evidence | Failure consequence | Test / sensitivity | |---|---|---|---|---| ## 8. REQUIRED OUTPUT 1. Research question and claim type. 2. Design choice and alternatives. 3. Variable / measurement plan. 4. Sampling plan. 5. Bias and validity map. 6. Analysis / preregistration outline. 7. Sensitivity requirements. 8. Inference boundaries. End with **Study Design Integrity Check** confirming that the proposed claim does not exceed what the design and data can support. This supports research planning and does not replace required ethics review, specialist methodological consultation or domain-specific standards.
分类与标签
学习教育、研究设计、问题质量审查、偏差评估、抽样计划