Research Question Identifiability and Design Feasibility Review

Public field comparison

Research Question Identifiability and Design Feasibility Review

Research Question Identifiability and Design Feasibility Review

Test whether a research question can be answered with the proposed design, population, variables, data and estimand, separating repairable gaps from fatal limitations.

Test whether a research question can be answered with the proposed design, population, variables, data and estimand, separating repairable gaps from fatal limitations.

Use this before committing to a study design or dataset when the researcher needs an evidence-based answerability review and a bounded redesign decision.

Use this before committing to a study design or dataset when the researcher needs an evidence-based answerability review and a bounded redesign decision.

Thesis Proposal Feasibility
Secondary Data Answerability Review
Research Design Triage

Thesis Proposal Feasibility
Secondary Data Answerability Review
Research Design Triage

research_question
proposed_design
available_data_and_population
constraints_and_decision_context

research_question
proposed_design
available_data_and_population
constraints_and_decision_context

Use this prompt with any capable AI assistant. Paste the research question and proposed design, then attach or quote a sanitized protocol, data dictionary, sampling plan and access constraints. Run the prompt before finalizing the method. Review every fatal or conditional limitation with the researcher and relevant methods or ethics advisers; do not submit identifiable participant data.

Use this prompt with any capable AI assistant. Paste the research question and proposed design, then attach or quote a sanitized protocol, data dictionary, sampling plan and access constraints. Run the prompt before finalizing the method. Review every fatal or conditional limitation with the researcher and relevant methods or ethics advisers; do not submit identifiable participant data.

A doctoral researcher has access to one cross-sectional institutional survey but wants to claim that a support programme caused retention. The review identifies the unavailable counterfactual, separates an answerable association question from the causal claim, and specifies the smallest design repair.

A doctoral researcher has access to one cross-sectional institutional survey but wants to claim that a support programme caused retention. The review identifies the absence of a defensible temporal and counterfactual comparison strategy, separates an answerable association question from the causal claim, and specifies the smallest credible design repair.

Expert

Expert

General AI

General AI

research design

research design

research
research design
research questions
research-methods
evidence-quality
causal-inference

research
research design
research questions
research-methods
evidence-quality
causal-inference

Research Question Identifiability and Feasibility Review

Research Question Identifiability and Feasibility Review

Test whether a research question is answerable with the proposed design, population, variables, data and estimand before committing study resources.

Test whether a research question is answerable with the proposed design, population, variables, data and estimand before committing study resources.

Prompt-body line comparison

Removed Added Unchanged context

 Review the proposed research question and design to determine what can actually be identified, estimated or learned from the available evidence.
 
 ## Context
 
 Research question:
 {{research_question}}
 
 Proposed design and estimand:
 {{proposed_design}}
 
 Available data, variables, population and sampling frame:
 {{available_data_and_population}}
 
 Constraints and decision context:
 {{constraints_and_decision_context}}
 
 ## Evidence rules
 
 - Separate supplied facts, researcher assumptions, design inferences, evidence conflicts, missing information and unresolved uncertainty.
 - Do not invent access to a population, variable, intervention, instrument, sample size, effect estimate, ethics approval or dataset field.
 - Distinguish a research question that is interesting from one that is answerable with the stated design.
 - Choose the methodological branch before testing feasibility. For quantitative questions, define the estimand and, where relevant, exposure or intervention, comparator, outcome and time horizon. For qualitative questions, define the phenomenon, context, participant or material boundary and intended interpretive claim. For mixed-method questions, define both branches and the point at which their evidence will be integrated.
 - Cite the supplied protocol, data dictionary, sampling document or source section for consequential conclusions when references are available.
−- Do not convert association into causation. When causal language is proposed, state the identification assumptions and whether the design can plausibly test them.
+- Do not convert association into causation. When causal language is proposed, state each identification assumption; distinguish assumptions that are empirically testable from those that are not; identify available diagnostics, falsification checks and sensitivity analyses; and assess whether each assumption is substantively defensible. Do not treat a passed diagnostic as proof of causal identification.
 - Minimize participant data. Use de-identified or aggregated descriptions whenever individual records are not essential, and do not reproduce direct identifiers or sensitive attributes in the output.
 
 ## Review method
 
 1. Restate the question as a decision-ready target and select `Quantitative`, `Qualitative`, or `Mixed method`. Apply estimands, comparators and statistical-identification tests only to the quantitative components for which they are relevant.
 2. Map every element of the question to observable evidence. For each construct, record its operational measure, source, timing, coverage, known validity limits and whether it is actually available.
 3. Follow the selected branch. For quantitative work, examine temporal order, selection, assignment, comparison structure, repeated measures, clustering, censoring, attrition and interference only where relevant. For qualitative work, examine sampling logic, access, contextual depth, data-generation method, analytic fit and reflexive limits. For mixed-method work, assess each branch and whether the proposed integration can answer the combined question.
−4. Assess answerability. For quantitative components, state which parameters are identified and which cannot be recovered. For qualitative components, state which interpretations the material can and cannot reasonably support. Classify every material assumption as `Design-guaranteed`, `Empirically diagnosable`, `Externally justified`, or `Untestable`, and cite its basis.
+4. Assess identification separately from estimation and precision. State which target quantities or claims are identified under the design and stated assumptions, which depend on additional or untestable assumptions, and which cannot be learned from the proposed evidence. Separately identify whether the available data and proposed methods can estimate the target with decision-useful precision.
 5. Examine population and sampling fit. Compare the target population with the sampling frame and observed sample; identify coverage, selection, non-response and generalizability limits.
 6. Check data feasibility. Review expected grain, linkage keys, missingness, measurement error, sample-size or information requirements and access constraints. Do not calculate power without the necessary assumptions.
 7. Classify every gap as repairable before data collection, repairable through a narrower claim, conditionally tolerable with disclosure, or fatal to the stated question.
 8. Compare the smallest credible options: retain the design, narrow the question, change measures, add data, change design, run a feasibility study, or stop. Preserve the researcher's authority over the final choice.
 
 ## Output contract: Answerability and Design Feasibility Record
 
 Return:
 
 1. **Question and method card**: exact question, selected methodological branch, target population or context, unit, material variables or phenomena, and claim type. Add an estimand, comparator and horizon only where relevant.
 2. **Evidence-to-construct map**: construct, required observation, supplied source, coverage, quality, timing and status.
−3. **Answerability assessment**: claim, design or method support, assumption classification, diagnostics or credibility checks available, alternative explanations or interpretations and confidence.
+3. **Identification, estimation and precision assessment**: claim, target quantity, identification status, required assumptions, assumption testability, available diagnostics or sensitivity analyses, estimation feasibility, precision evidence, alternative explanations and confidence.
 4. **Feasibility register**: gap, evidence, impact, classification as repairable/conditional/fatal, smallest repair and owner. Use `Unassigned` when no owner is supplied rather than inventing one.
 5. **Option comparison**: each viable design or scope option, what it can answer, what it cannot answer, cost or burden evidence and unresolved risk.
 6. **Researcher decision note**: recommend `Proceed`, `Proceed after repair`, `Narrow the question`, `Redesign`, or `Not currently answerable`, with the evidence supporting the recommendation. Label it advisory.
 
 ## Verification and completion
 
 Before calling the review complete, verify that every material term maps to appropriate evidence for the selected methodological branch; quantitative estimands and comparators are explicit only where relevant; qualitative context and interpretive limits are explicit; mixed-method integration is justified; population and timing align; assumptions use the required classifications; missing data are visible; fatal limitations are not disguised as caveats; and the proposed next step can materially reduce uncertainty.
 
 If critical design, data or population information is missing, return a provisional map and a focused evidence request instead of a final feasibility conclusion. Refuse requests to certify validity, ethics approval or causal identification without the corresponding evidence and authorized review.