🔬 Research
Research Methodology Prompt: Design, Sampling and Analysis
Stress-test your methodology: design fit for each research question, sampling and sample size math, instruments, analysis plan, validity threats and ethics.
0Reviews
Prompt
Act as a research methods advisor who coaches graduate students and early-career researchers on their methodology. You do not write the chapter for me; you check whether my design can answer my questions, show the reasoning, do the arithmetic, and tell me what my supervisor or ethics board will ask. You never invent citations, instruments, scores or rules; if something needs a source, you say what kind of source to look for. Inputs: - My research questions or hypotheses, numbered: [ResearchQuestions] - The design I am planning (survey, interviews, experiment, mixed methods, secondary data): [PlannedDesign] - Population, setting and how I can reach people: [PopulationAccess] - Sample size I am considering and any expected response rate: [SampleSizeInfo] - Instruments or measures I plan to use, with their source if I know it: [Instruments] - Analysis I am planning: [AnalysisPlan] - Constraints (time, budget, ethics, access): [Constraints] - My program's requirements for the methodology chapter: [ProgramRequirements] Generate: 1. Design fit table: each ResearchQuestion, what kind of answer it needs (describe, compare, explain, explore), whether PlannedDesign can give it, and why. 2. Sampling: the sampling method that fits PopulationAccess, its main bias risk, and sample size arithmetic shown step by step. Say which formula you used and what it does and does not cover (precision for a proportion is not power for a test). 3. Instruments: for each, what it measures, what to confirm in the original source (version, scoring, permission), and any gap between the measure and the question. 4. Analysis plan: the analysis for each ResearchQuestion, assumptions to check, and what result would answer it. 5. Validity and trustworthiness: the three biggest threats, each with a mitigation I can actually do under Constraints. 6. Ethics checklist: consent, data storage, risks to participants, and what the ethics board will likely ask. 7. Supervisor questions: five questions a supervisor or committee will ask, with what I need ready for each. Rules: show every calculation. Mark anything to verify as [CHECK]. Do not write my chapter text; give me the reasoning and the decisions.
Instructions
Number your research questions and paste them exactly; most methodology problems show up when one question needs a different design. Variation: change PlannedDesign to an experiment and ask for randomization and power checks instead. Tip: run the prompt again after your supervisor meeting with their comments added as Constraints. Never let the model write your chapter or add citations; use it to check your reasoning and look up every source yourself. Follow your program's rules on AI use.
Generated Output
This image was generated using the prompt above.

Examples
Example Input
ResearchQuestions: RQ1 What proportion of community garden members aged 65 and over report frequent loneliness? RQ2 Is time spent gardening associated with loneliness scores, after accounting for living alone? RQ3 How do members describe the role of the garden in their social life? PlannedDesign: mixed methods: online and paper survey, then 12 semi-structured interviews. PopulationAccess: the city's community garden network has 2,400 registered members aged 65 and over; it will send my survey link and paper copies. SampleSizeInfo: I need precision of plus or minus 5 percentage points for RQ1 at 95% confidence; I expect about a 30% response rate. Instruments: 3-item UCLA Loneliness Scale; hours gardened per week (self-report); living alone (yes or no). AnalysisPlan: descriptive statistics for RQ1; correlation for RQ2; thematic analysis for RQ3. Constraints: one semester for data collection, no budget for incentives. ProgramRequirements: methodology chapter must justify design, sampling, instruments, analysis and ethics.
Example Output
1. DESIGN FIT | RQ | Needs | Can the design answer it? | | RQ1 | Describe a proportion | Yes, from the survey, if the sample is large enough (see 2). | | RQ2 | Explain an association with a control variable | Partly. A correlation cannot adjust for living alone. Use multiple linear regression with living alone as a covariate. Cross-sectional data shows association, not cause. | | RQ3 | Explore meaning | Yes, interviews with thematic analysis. | 2. SAMPLING AND SAMPLE SIZE Method: census invitation to all 2,400 members through the network; respondents self-select, so the main risk is that the most socially active members answer most. Formula for RQ1 (precision of a proportion, p = 0.5 as the most cautious value): n0 = 1.96 x 1.96 x 0.5 x 0.5 / (0.05 x 0.05) = 384.16 Finite population correction for N = 2,400: n = 384.16 / (1 + (384.16 - 1) / 2,400) = 331.3, round up to 332 completed surveys. Invitations at a 30% response rate: 332 / 0.30 = 1,106.7, so about 1,107. You plan to invite all 2,400, which is enough if the 30% holds (0.30 x 2,400 = 720 expected). What this does not cover: power for RQ2. Run a power analysis for regression with two predictors (for example in G*Power) and state the effect size you assume. [CHECK] Interviews: 12 is a reasonable plan for thematic analysis; justify it by the aim (depth, not representativeness) and report when no new codes appeared. 3. INSTRUMENTS - 3-item UCLA Loneliness Scale: confirm the version, the response options, the scoring range and any cut-off for "frequent loneliness" in the original source. RQ1 needs that cut-off defined before data collection. [CHECK] - Hours gardened per week: self-report; ask for a typical week in the growing season and say so. - Living alone: yes or no, fine as a covariate. 4. ANALYSIS PLAN RQ1: proportion above the defined cut-off with a 95% confidence interval. RQ2: multiple linear regression, loneliness score on hours gardened plus living alone; check linearity, residual normality and outliers. RQ3: reflexive thematic analysis; keep an audit trail of codes. 5. VALIDITY THREATS - Nonresponse bias: compare respondents to the network's age and gender totals if it can share them. - Mode effects (online vs paper): record mode and test whether scores differ by it. - Seasonality: one semester; state the months of data collection as a limitation. 6. ETHICS Consent wording for older adults on paper and online; how paper forms are stored; a support line on the information sheet, since loneliness questions can be upsetting; no incentives, so state that participation is voluntary with no benefit. 7. SUPERVISOR QUESTIONS Why regression and not correlation for RQ2? What is your loneliness cut-off and its source? How will you reduce nonresponse bias without incentives? How will you pick the 12 interviewees? How do the survey and interview results connect?