design-a-research-method

# design-a-research-method Choose the least burdensome method capable of answering the bounded question. ## When to use - Use for observational, experimental, qualitative, quantitative, mixed, archival, computational, or design research. - Do not begin collection involving people, sensitive data, hazards, or protected sites before required approval. ## Procedure 1. State the question, target population or system, unit of analysis, concepts, and decision the evidence should inform. 2. Define the claim type: descriptive, comparative, relational, causal, predictive, interpretive, or methodological. 3. Map each concept to observable evidence and identify likely confounders, biases, and alternative explanations. 4. Compare candidate designs for validity, ethics, burden, access, cost, timing, and analytic fit. 5. Select sampling or case-selection logic, inclusion and exclusion rules, comparison conditions, and stopping criteria. 6. Specify measures, instruments, procedures, timing, quality controls, and pilot tests. 7. Predefine analysis, missing-data treatment, uncertainty, subgroup logic, and sensitivity checks where appropriate. 8. Plan consent, privacy, safety, data governance, accessibility, and researcher positionality. 9. Run a feasibility pilot without using unapproved pilot data as final evidence. 10. Document limitations and the claims the design cannot support. ## Worked example A project asks whether a tutoring program improves grades. The method distinguishes a causal question from a satisfaction survey, identifies selection bias, and compares randomized, matched, and interrupted-time designs. Constraints lead to a prospective matched design with preregistered outcomes and explicit residual limitations. ## Done - A method design document links question, claim type, sampling, measures, procedure, analysis, governance, and feasibility - A design review verifies that planned evidence supports only the stated claims and records unresolved bias plus required approvals