RW Research Design

Startup

1. Read

   references/standalone.md

   ,

   references/method.md

   , and

   references/standards.md

   .
2. Read the materials provided by the user in this round; if no materials are provided, only complete the parts allowed by current evidence.
3. Retrieve

   references/atoms.jsonl

   when rule judgment is needed; read

   references/cases.md

   when encountering similar tasks.
4. Read

   references/domain-guide.md

   when selecting methods or tools, and use

   assets/worksheet.md

   to organize deliverables.
5. Read

   references/acceptance.md

   .

   references/behavior-tests.json

   is only used for testing and not as factual information for user tasks.
6. When current literature, APIs, report specifications, and journal requirements may change, open the official links in

   references/source-map.md

   for verification and record the date.

Work Stages

1. Write down the main questions, target estimators, or phenomena that need explanation.
2. Define constructs, variables, measurement time points, data sources, and target populations.
3. Compare the suitability of experimental, observational, diagnostic, predictive, qualitative, mixed-method, and economic designs.
4. Identify control, confounding, selection, measurement, missing, time, and implementation biases.
5. Preset primary analysis, sensitivity analysis, sample justification, and failure criteria.
6. Output the preferred design, alternative designs, resource constraints, ethical considerations, and report specifications.

Operating Rules

• Research design starts from the problem and target estimator, not from software functions.
• Report specifications are used to check transparency and cannot replace design rationality.
• Randomization can handle known and unknown confounding, but cannot automatically solve attrition, measurement, and execution biases.
• Observational studies must clarify temporal order, confounding structure, and target populations.
• Diagnostic studies must clarify the index test, reference standard, threshold, and usage scenario.
• Predictive studies must distinguish between development, internal validation, external validation, update, and impact evaluation.
• Qualitative studies must align sampling, context, researcher position, and analysis methods with the problem.
• Mixed-method designs must explain when, why, and how the two types of data are integrated.
• Sample size justification must correspond to the main question, design, and analysis, not just provide a number.
• Primary and secondary analyses, subgroup and exploratory analyses must be distinguished in advance.
• Feasibility constraints should trigger alternative designs, not silently reduce the strength of conclusions.
• The design must specify what results would weaken the main interpretation.

Output

• Research design and target estimator.
• Variables, samples, biases, and analysis plan.
• Alternative designs, failure criteria, and pre-execution gaps.

Stop Conditions

• When existing data cannot answer the question, clearly mark it as unanswerable.
• When ethical, licensing, or data governance information is missing, do not write the plan as immediately executable.
• Do not package observational associations as evidence from randomized trials.
• Do not provide conclusions that replace the approval of statisticians, ethics committees, or domain experts.

Independent Operation

• By default, do not read private workspaces, personal corpus directories, or preset research-lab.
• Files, text, links, and data provided by the user are inputs for the current task, not installation dependencies.
• When the network is unavailable, continue using stable methods within the Skill; mark parts requiring current facts as pending verification.
• Can connect with other research Skills in this package if they exist; when installed alone, directly return instructions for the next step without stopping the current task.

Source Discipline

• Separate user materials, public sources, current inferences, and unknown matters.
• Report specifications only check report transparency and do not automatically prove design quality.
• Public source abstracts are stored within the Skill; return to official pages when version, cost, policy, API, or journal status is needed.
• Do not generate non-existent papers, data, DOIs, tool operation results, or journal requirements.