Your own study: from a question to a report

Plan a study at one scale, write a one-page proposal, carry it out, and report it in a paper, a poster and a rubric sheet

The research capstone (CR) of Energy Systems, Step by Step: take a question card from the question studio, choose a track (model-based, empirical or mixed) and a scale, plan the stages from question to report with their milestones and studios, write a one-page proposal, carry out the study in 20 to 40 hours, and judge it against the course rubric with weak and strong examples of practice.
  • Researcher · L4
  • Capstone
  • 60 min guided · about 30 h practice
  • After R10 and R2 and R5 and R6 and R7 and R8 and R9 and CK-M

The brief

Design, carry out and report a study of your own at one scale: a question you chose, a method you can defend, and a result someone else can rerun.

The course's research capstone. It certifies nothing; it gives you the evidence a mentor needs to judge whether you are ready for supervised research.

  1. Hand in: a one-page proposal, a reproducible repository (or an audit trail for qualitative work), a six-page paper, a poster and your rubric sheet.
  2. From: a question card from the question studio, one scale route (its model lab and capstone), the research studios and CK-M.
  3. Done when: every stage below meets its own check, CK-R's mock review is answered, and your self-check reaches Proficient on every competence.

About 20 to 40 hours of your own work, CK-R's four among them: about an hour on this page, then the study. A mentor's review and a peer-review swap are optional, and help most.

Three tracks, one rubric

Model-based

Optimization or simulation. A model of the system that answers what to do, or what would happen.

Questions
Prescriptive, predictive or explanatory
Rerun
Reruns from a clean clone; checked against a hand calculation or a published benchmark, and validated against real data.
Lessons
M1, M3, B3

Empirical

Data analysis, or a reproduction or replication of a published result. An analysis of observed data; or a published result rerun with its own data and code (a computational reproduction), or tested on new data (a replication).

Questions
Descriptive, explanatory, evaluative or predictive
Rerun
Reruns from a clean clone, raw data to result; a reproduction or a replication reports how close it came to the published number.
Lessons
B2, M14, R4

Mixed

A model or data analysis with a stakeholder or qualitative part. Numbers and people's reasons, with a plan for how each informs the other.

Questions
Explanatory, evaluative or prescriptive
Rerun
A clean-clone rerun for the numbers; an audit trail from raw notes to themes for the qualitative part.
Lessons
R9, A13, R3

Every track is judged on the same nine competences. A fully qualitative study is allowed with a mentor; the course teaches qualitative methods at awareness level only.

The five scale routes: a model lab and a capstone each
Scale Model lab Capstone
S0 device and plant M4 C0
S1 microgrid and community M5 C1
S2 regional grid M2 C2
S3 city and region M12 C3
S4 nation M13 C4

Plan your study

The model-based track: about 30 hours (20 to 40). At 6 hours a week, about 5 weeks (4 to 7).

  1. Question · 1.5 h
  2. Literature · 5 h
  3. Method · 1.5 h
  4. Data · 2 h
  5. Build and validate · 6.5 h
  6. Uncertainty and accounting · 2 h
  7. Mock review and rerun · 4 h
  8. Paper and poster · 6 h
  9. Integrity and release · 1.5 h

Underlined: a stage that ends on a milestone (3, the one-page proposal; 7, CK-R; 8, the paper and the poster; 9, the rubric sheet).

Your choice Starting plan Range
Track Model-based one of: Model-based, Empirical, Mixed
Scale S1 microgrid and community one of: S0 device and plant, S1 microgrid and community, S2 regional grid, S3 city and region, S4 nation
Method family Optimization one of R3's nine method families
Alternatives Simulation, Agent-based modelling any of the nine families
Uncertainty treatment Scenarios one of: Scenarios, Global sensitivity analysis, Monte Carlo, Stochastic or robust optimization, Intervals and robustness checks, Triangulation, One input at a time
Hours a week 6 h 2 to 20 h, in steps of 1 h
Mock review by A peer (the swap) one of: A mentor, A peer (the swap), Yourself, after a day away
Stages done none any of the nine stages

With JavaScript on, you pick a track, a scale, a method and your hours a week, and the page sizes each stage.

Without it, plan with the table below.

Stage Do, then check Studios Hours: Model-based · Empirical · Mixed
1. Question Turn your question card into one you can answer at your scale, with its boundary.
  • Your card passes its eight checks.
  • You can say what answer would change your audience's decision.
 R1, R10, R3 1–2 · 1–2 · 1–2
2. Literature Search open databases, snowball from a review, and map what is known.
  • A concept matrix of at least 12 sources.
  • A gap stated with evidence, from a search a reviewer can repeat.
 R2 4–6 · 4–6 · 4–6
3. Method Choose a method family, defend it, and write the one-page proposal.
  • The method defended against two alternatives, one from another family.
  • Milestone: the proposal.
 R3, your route's model lab 1–2 · 1–2 · 1–2
4. Data Get, clean and document your data, or plan and hold your interviews.
  • A data note: each source with its date and licence.
  • A cleaning log; for interviews, consent forms and a guide.
 R4, B2 1–3 · 3–8 · 2–6
5. Build and validate Build the model or analysis under version control, verify it, and validate one result.
  • It reruns from a clean clone (or a complete audit trail).
  • A verification log.
  • One validation against real data or an independent reference.
 R5, CK-M, K8, your route's model lab 3–10 · 2–7 · 2–7
6. Uncertainty and accounting Classify the uncertainties, treat each one, and put cost and carbon on one boundary.
  • Uncertainties classified.
  • Scenarios or a global sensitivity analysis; results as ranges.
  • Cost and emissions on one boundary where you report both.
 R6, R7, M7, M8 1–3 · 1–3 · 1–3
7. Mock review and rerun Get a written mock review of your question and method, answer it, revise, and rerun your main result.
  • Every major concern answered.
  • A change log.
  • A clean-clone rerun within your tolerance, or the difference explained.
 CK-R 4 · 4 · 4
8. Paper and poster Write the paper and the poster, with each result's limits beside it.
  • A six-page paper with one key figure.
  • A limits paragraph and a 150-word public summary.
  • A poster.
 R8, your route's capstone 4–8 · 3–6 · 3–6
9. Integrity and release Check licences, credit and conflicts; treat the people your study is about fairly; release what you may.
  • Licences checked; credit and conflicts stated.
  • The people the study is about treated fairly.
  • Data and code released with documentation where allowed.
  • Your rubric sheet.
 R9 1–2 · 1–2 · 2–4
All The study, from question to report. The studios above 20–40 · 20–40 · 20–40

The hours are the curriculum's planning ranges for a first study, not measurements; the weeks follow from your hours a week. Method families as in R3. Your plan stays in this browser and in its link.

Write your one-page proposal

  1. Title and plan link: the plan's link from the planner above, and the date
  2. Why it matters: Who would act on the answer, and what answer would change their decision.
  3. The plan: your question card's eight parts; the track, the scale route, the method and its alternatives, the mock review, the time and the milestones
  4. Assumptions: the hours a week the weeks rest on
  5. Sensitivity: the uncertainty treatment you plan
  6. What is known, and the gap: The three closest studies and what none of them answers; the size of your concept matrix and the date of your search (R2).
  7. Why this method, not the alternatives: What your method answers that the two alternatives cannot (name a same-family one exactly), and one limit you will mitigate (R3).
  8. Data and licences: Each source, its date and its licence; for interviews, who you will ask and how you will ask for consent (R4, R9).
  9. How you will check it: One verification check, one validation against real data or an independent reference, and the tolerance you set now (R5).
  10. What the study leaves out: The parts, people and years outside your boundary, and why that does not change the answer to your question.
  11. Integrity: Credit, conflicts, consent, and who the study affects (R9).

With JavaScript on, your question card and your plan fill in the proposal's table, and the checks say which parts are in place. Without it, write the proposal under these headings, on one page.

Go deeper · ResearcherA mentor's review and the peer-review swap
  • A mentor reads three things: the proposal, the validation evidence and the final study. Send each with its rubric self-rating.
  • The swap: two learners exchange proposals and review each other’s with CK-R’s review form, within a week; then answer point by point, and swap again for the paper’s draft.
  • The reviewer’s duties: keep the other proposal confidential, do not use its ideas without permission, and declare any conflict (ALLEA 2023, §2.8).
  • Credit: name your reviewers in the paper’s acknowledgements.

Check your study

Framing: the question

Stage 1, question · Prairie City's 2023 weatherization programme: did weatherized homes use less gas than homes on its waiting list (empirical, evaluative, S3). Write the study's question as a card. Both fragments are illustrative; the study is fictional and its numbers are assumed, not computed.

Weak · Developing

Question: Is weatherization good for Prairie City? [1]

  1. Falls short: No type, no audience, no metric, no comparison: nobody could say when it is answered.

Strong · Strong

Question: Did homes weatherized in 2023 use less gas [1] in the next two winters than similar homes still on the waiting list? [2] Type: evaluative. Audience: the city's housing office, which sets next year's budget and would cut the programme if the saving is under 5%. [3] Metric: winter gas use per home, kWh, adjusted for the weather. Scale: S3; boundary: the homes' gas, not their electricity. Wrong if: the waiting-list homes' use fell as much. [4]

  1. Works: A metric; its units follow.
  2. Works: The counterfactual: homes that applied too.
  3. Works: Who acts on the answer, and the answer that would change the decision.
  4. Works: A falsification statement.

Positioning: the gap

Stage 2, literature · Prairie City's 2023 weatherization programme: did weatherized homes use less gas than homes on its waiting list (empirical, evaluative, S3). State what is known and the gap. Both fragments are illustrative; the study is fictional and its numbers are assumed, not computed.

Weak · Developing

Gap: few studies look at weatherization in Prairie City, so this study is novel. [1]

  1. Falls short: Few studies is not evidence of a gap, or of novelty: the topic may be unimportant, or the search poor.

Strong · Strong

Known: of the 14 studies in my concept matrix, 9 use billing data and 3 compare with a waiting list; [1] none of the 3 is in a city with winters this cold. [2] Search: OpenAlex, with the string and the date in appendix A; two more studies found, both in milder climates. [3] Closest: a waiting-list study in a milder city; this one adds cold winters and a second winter. [4]

  1. Works: The map behind the claim: a matrix of more than 12 sources.
  2. Works: The empty cell, stated precisely.
  3. Works: A dated search a reviewer can repeat.
  4. Works: The closest prior work, engaged directly.

Method fit: the choice

Stage 3, method · A co-op's farm feeder: how much battery, and which tariff, keep the summer evening peak under the feeder's limit at least cost (model-based, prescriptive, S1). Choose the method and defend it. Both fragments are illustrative; the study is fictional and its numbers are assumed, not computed.

Weak · Developing

Method: optimization, because I used it in M2. [1]

  1. Falls short: Chosen by habit.

Strong · Strong

Method: a linear program of the feeder's summer evenings: the question asks what to do, under stated rules. [1] Against simulation: it tests the designs it is given, but does not search for the cheapest. [2] Against agent-based modelling: it would capture how members respond, but needs behaviour data the co-op does not have. [3] Limit: perfect foresight flatters the battery; I rerun with a day-ahead forecast. [4]

  1. Works: The family fits the question's type (prescriptive); a linear program is one kind of optimization.
  2. Works: One alternative, and why not.
  3. Works: A second alternative, from another family.
  4. Works: A limit of the chosen method, stated and partly mitigated (perfect foresight).

Trustworthy analysis: the log

Stage 5, build and validate · A co-op's farm feeder: how much battery, and which tariff, keep the summer evening peak under the feeder's limit at least cost (model-based, prescriptive, S1). Write an entry of the verification log. Both fragments are illustrative; the study is fictional and its numbers are assumed, not computed.

Weak · Developing

Log: checked the model; results look reasonable. [1]

  1. Falls short: No check named, no reference, no tolerance: nobody can repeat it.

Proficient · Proficient

Balance: holds in every hour (largest error 0.000 kW). [1] Benchmark: with no battery, the bill equals the hand calculation to the cent. [2] Validation: the feeder's metered 2025 evening peak, 2.81 MW, against the model's 2.76 MW: 1.8% low, inside the 5% set beforehand. [3] Rerun: from a clean clone, the same result. [4]

  1. Works: A verification check, named, with its result.
  2. Works: An independent reference, worked by hand.
  3. Works: Real data, with a tolerance set before looking.
  4. Works: Reproducible from the repository alone (clean clone). Strong would add a rerun by someone else.

Uncertainty: the result

Stage 6, uncertainty and accounting · The Plains Grid in a five-day heat wave: how often the reserve falls short with more air conditioning (model-based, predictive, S2). Report the main result. Both fragments are illustrative; the study is fictional and its numbers are assumed, not computed.

Weak · Developing

Result: the reserve falls short in 4 hours of the heat wave. [1]

  1. Falls short: A single number, as if certain.

Proficient · Proficient

Result: in 90% of 1,000 Monte Carlo runs the reserve falls short in 1 to 9 hours of the heat wave, against 0 to 2 hours with today's air conditioning. [1] Drivers: air-conditioning growth moves it most, then the gas fleet's outage rate. [2] Kept apart: we treat a heat wave hotter than any on record as a stress scenario, because we cannot defend a probability for it. [3]

  1. Works: A range, how it was made, and what it is compared with.
  2. Works: The uncertainties that matter, ranked. Strong would add what knowing them better is worth to the decision.
  3. Works: Deep uncertainty kept out of the probabilities: the evidence, not the record alone, decides it.

Consistent accounting: cost and carbon

Stage 6, uncertainty and accounting · A co-op's farm feeder: how much battery, and which tariff, keep the summer evening peak under the feeder's limit at least cost (model-based, prescriptive, S1). Report cost and CO₂. Both fragments are illustrative; the study is fictional and its numbers are assumed, not computed.

Weak · Developing

Result: the battery saves money and cuts emissions a lot. [1]

  1. Falls short: Whose money, on what boundary, against what?

Proficient · Proficient

Cost: against no battery, the co-op's power bill falls $250,000 a year; less the battery's annualized cost, $160,000, its yearly cost falls $90,000 (2025 dollars). [1] CO₂: 120 t a year less, location-based with the grid's hourly average factors, on the same boundary: the co-op's purchases. [2] How: it buys 2,000 MWh at night (0.39 t CO₂ per MWh) to deliver 1,800 MWh in the evening (0.50 t per MWh): 1,800 × 0.50 − 2,000 × 0.39 = 900 − 780 = 120 t. [3] Who: members pay for the battery through rates; farms on the feeder gain most. [4]

  1. Works: Whose money, against what, in which dollars: the bill's fall less the battery's yearly cost.
  2. Works: One boundary for cost and carbon; the emissions question named (location-based).
  3. Works: The arithmetic, its 10% losses included. With one yearly factor for every hour, the 200 MWh lost would add CO₂ instead.
  4. Works: Who pays and who benefits, named.

Realism: the inputs

Stage 4, data · A co-op's farm feeder: how much battery, and which tariff, keep the summer evening peak under the feeder's limit at least cost (model-based, prescriptive, S1). Say what the key inputs are, and what the study leaves out. Both fragments are illustrative; the study is fictional and its numbers are assumed, not computed.

Weak · Developing

Inputs: the co-op's new 2 MW battery is installed in 2027. [1]

  1. Falls short: An option written as if it were a project that exists.

Proficient · Proficient

Battery: a design option; the co-op has decided nothing. [1] Tariff: the co-op's published 2025 tariff, dated. [2] Leaves out: the feeder's voltage limits, battery wear after year 10, and how members respond to the tariff. [3]

  1. Works: Evidence status: an option, not an announced or built project.
  2. Works: A primary document for a key input.
  3. Works: A list of what the analysis leaves out.

Communication: the public summary

Stage 8, paper and poster · Prairie City's 2023 weatherization programme: did weatherized homes use less gas than homes on its waiting list (empirical, evaluative, S3). Write the core of the 150-word public summary. Both fragments are illustrative; the study is fictional and its numbers are assumed, not computed.

Weak · Developing

Summary: our study proves weatherization works and should be extended to every home. [1]

  1. Falls short: Proves, works, every home: no size, no limits, and a policy the study did not test.

Proficient · Proficient

Finding: weatherized homes used 9% to 15% less gas in the two winters after than similar homes still on the waiting list. [1] Limits: one city, two winters, homes that applied; gas use, not comfort or health. [2] For the housing office: the result holds for homes like those that applied; homes that never apply may differ. [3]

  1. Works: The finding, with its size and its comparison.
  2. Works: The limits beside the result.
  3. Works: A reading that is hard to misread.

Integrity: the statement

Stage 9, integrity and release · A co-op's community solar: why some farms joined and others did not, from sign-up records and 12 interviews (mixed, explanatory, S1). Write the integrity statement. Both fragments are illustrative; the study is fictional and its numbers are assumed, not computed.

Weak · Developing

Ethics: no issues: the data are public. [1]

  1. Falls short: The interviews and the co-op's records are not public: consent and licences unchecked.

Strong · Strong

Consent: the 12 interviewees signed a consent form, and the university's ethics board approved the interviews. [1] Data: the co-op's sign-up records under a data-sharing agreement, names removed; I release the code and a synthetic copy, not the records. [2] Credit: co-op staff in the acknowledgements; no funding and no conflicts. [3] Findings: the co-op saw them before publication, with no veto. [4]

  1. Works: Consent and ethics review for the people the study is about.
  2. Works: The licence checked; released where allowed (a synthetic copy).
  3. Works: Credit and conflicts stated.
  4. Works: Stakeholders treated fairly, without control of the results.
Competence Developing Proficient (the bar) Strong
Framing
Look for: Your proposal's question card passes its eight checks and says what answer would change your audience's decision.		 Question vague or untyped; no audience or counterfactual. Typed question with audience or decision-maker, metric with units, counterfactual, scale and boundary; one falsification statement. Also explains why the question matters to its audience and what answer would change a decision.
Positioning
Look for: Your concept matrix of 12 sources or more, and a dated search behind the gap.		 A few sources, no map. Concept matrix of at least 12 sources; a stated gap with evidence. The gap survives a search by a reviewer; closest prior work engaged directly.
Method fit
Look for: Your method defended against the two alternatives in your plan, and one limit you mitigated.		 Method chosen by habit. Method defended against two alternatives, one of them a different family. Limits of the chosen method stated and partly mitigated.
Trustworthy analysis
Look for: A clean-clone rerun (CK-R), a verification log and one validation, named in the paper.		 Results cannot be rerun. Reruns from a clean clone (or a complete audit trail); verification log; one validation against real data or an independent reference. Independent reproduction by someone else; validation on data not used to build the analysis.
Uncertainty
Look for: The main result as a range, with how the range was made.		 Single numbers. Uncertainties classified; scenario or global sensitivity analysis; results as ranges. Decision-relevant uncertainty identified and its value discussed.
Consistent accounting
Look for: Cost and emissions on one boundary where you report them, the emissions question named, and who pays and who benefits.		 Costs and emissions on different boundaries. Cost and emissions on one boundary; emissions question named; who pays and who benefits named. Distributional and environmental side effects estimated with stated assumptions.
Realism
Look for: The evidence status of your key inputs, and a list of what the study leaves out.		 Assumes announced projects exist. Evidence status of key inputs; a 'what the analysis leaves out' list. Realism checked with a practitioner or primary documents.
Communication
Look for: The six-page paper, its key figure, the limits paragraph and the 150-word public summary.		 Results without limits. Six-page paper; one key figure; a limits paragraph; a 150-word public summary. A briefing a decision-maker could act on without misreading.
Integrity
Look for: Your integrity statement: licences, credit, conflicts and consent; what you released.		 Licences or credit unchecked. Licences checked, credit and conflicts stated, stakeholders treated fairly. Data and code released with documentation where allowed.

Two statuses you can claim, in your own words

Status Requires Evidence you can show
Completed Research Studio The core path, the checkpoints CK-F, CK-A and CK-M, and a submitted CR with a self-assessment. The CR repository and report, the rubric sheet.
Demonstrated supervised-project readiness at scale Sx The above, plus the Sx route, CK-R, and Proficient or better on all nine rubric competences. A mentor's review of the proposal, the validation evidence and the final study is optional but recommended. The same, plus the CK-R review and response, and the mentor's comments if any.

The course issues no certificate or badge. A web course cannot verify who did the work or replace a mentor's judgement: it gives you a published standard and the evidence to show against it.

Numbers, sources and what’s next

Numbers used on this page

Number What it means Source
20 to 40 hours The study, CK-R's four hours among them: each track's stages sum to this range. The course's planning range
12 sources The least a concept matrix holds at the rubric's Proficient level. The course rubric
6 pages, 150 words The paper, and its public summary, at the rubric's Proficient level. The course rubric
Fragments Every number in the weak and strong fragments: assumed for the example, not computed. The cost and CO₂ fragment shows its arithmetic. Illustrative

Next

CK-R · Researcher checkpoint: defend your question and method, then revise

Sources and further reading

Spotted an error, or stuck? Write to hlli@ksu.edu. Corrections are made in place and dated here.

Text licensed CC BY 4.0, code MIT: teachers and students may reuse and adapt it, with credit to Honglin Li.