Chapter 7 — Dissertation writing and limitations
Lesson map
Lesson 07 published. Map, example, EX039.
flowchart TD
N0["Learning outcomes"]
N1["7.1 Structure"]
N2["7.2 Limitations"]
N3["7.3 Findings versus recommendations"]
N4["7.4 Recommendations"]
N5["7.5 Figures"]
N6["7.6 Writing clearly"]
N7["Appendices"]
N0 --> N1 --> N2 --> N3 --> N4 --> N5 --> N6 --> N7
N7 --> EX["Worked example"]
EX --> Q["Exercise EX039"]
Logical drawing
flowchart LR
N0["Learning outcomes"]
N1["7.1 Structure"]
N2["7.2 Limitations"]
N3["7.3 Findings versus recommendations"]
N4["7.4 Recommendations"]
N5["7.5 Figures"]
N6["7.6 Writing clearly"]
N7["Appendices"]
N0 --> N1 --> N2 --> N3 --> N4 --> N5 --> N6 --> N7 --> Q["EX039"]
[Learning outcomes]
|
v
[7.1 Structure]
|
v
[7.2 Limitations]
|
v
[7.3 Findings versus recommendations]
|
v
[7.4 Recommendations]
|
v
[7.5 Figures]
|
v
[7.6 Writing clearly]
|
v
[Appendices]
|
v
[Worked example] --> [Exercise EX039]
Learning outcomes
You will (1) write the dissertation clearly; (2) present limitations honestly; (3) separate findings from recommendations; (4) prepare figures and appendices.
7.1 Structure
Introduction and question; literature review with the search method; design and ethics; results by explanation; discussion; limitations; recommendations; appendices with search log, analysis scripts and the synthetic generator. The structure mirrors the study's logic: each explanation gets its own results heading, with its prediction stated first and the evidence after. Findings and recommendations are kept in separate chapters. Appendices let an examiner rerun every table from the archive without contacting Lucas. A short abstract states the question, the three explanations and the main results with their limits.
7.2 Limitations
One organisation and one year; origin labels depend on logging accuracy; intended public resources depend on a register that may be incomplete; the study measures exposure, not whether data were accessed. Each limitation is tied to a claim. One organisation limits how far the rates generalise, though the mechanisms may transfer; logging accuracy matters for the origin comparison; the register matters for the exposure measure; and the study cannot say whether any exposed data were read. Lucas states these where the related results appear, not only in a final list.
7.3 Findings versus recommendations
Findings say what the data show; recommendations say what Larkspur might do, labelled as Lucas's judgement. Mixing them would overstate the evidence. The dissertation marks the boundary in its wording: results use "in this dataset", and recommendations begin with phrases such as "Lucas recommends". Dr Tanaka's comment on the draft sentence claiming the recommendations "will prevent exposures" is an example of this boundary being enforced. The revised sentence names a test Larkspur could run. Each recommendation names the explanation and table it rests on, so Larkspur can judge how strong the evidence is for that change.
7.4 Recommendations
Block public access at account level by default with approved exceptions, as provider guidance suggests; fix the two templates and add policy checks to IaC review; reduce storage administrators; route urgent data requests through a pre-approved, private export process. Each recommendation answers one explanation: account-level blocking and the private export process for console changes, template fixes and policy checks for IaC, and fewer storage administrators for identity breadth. Lucas labels them as his judgement based on the findings, and suggests repeating the measurement to see whether they work.
7.5 Figures
One figure per explanation shows its prediction against the data, using the same scale. Using the same scale prevents a small difference from looking dramatic in one figure and a large one looking minor in another. Each figure also shows the counts behind the rates. A reader comparing the three explanations can then see at a glance which predictions fit the data best. Figures are produced by the archived scripts, so a rerun reproduces them exactly, and each caption states the dataset and period it covers.
7.6 Writing clearly
Short paragraphs, defined terms and one claim per paragraph. Dr Tanaka asks Lucas to remove phrases like "clearly" and "proves", which overstate observational results. Each defined term, such as unintended public access, is used the same way throughout, so readers do not have to guess whether exposure means something different in one chapter. A glossary at the front lists the terms with their definitions. Dr Tanaka reads one chapter at a time and marks any paragraph that makes more than one claim, which Lucas then splits.
Appendices
The appendices let readers check everything without cluttering the main text. They contain the full search log with screening decisions, the pre-registration and its dated amendments, the analysis scripts with a short guide to rerunning them, the synthetic data generator and a table mapping each recommendation to the finding that supports it. Dr Tanaka asks Lucas to test the guide by having a classmate follow it; the classmate reproduces every figure on the first attempt.
Word limits
Lucas keeps within the word limit by moving detail to appendices.
Worked example
In draft review, Dr Tanaka marks a sentence claiming the recommendations "will prevent exposures". Lucas changes it to "are expected to reduce exposures, which Larkspur could test by repeating the measurement after a year".
Synthetic case. Text and outcomes are synthetic.
Errors and limits
Recommendations framed as findings mislead readers about certainty.
Exercises
- EX039 — Limitations. Name three limitations. Rubric: 3 = one org/year, logging accuracy, register, no access data; 2 = two; 1 = one; 0 = none.
- EX040 — Findings vs recommendations. How do findings differ from recommendations? Rubric: 3 = data vs judgement; 2 = partial; 1 = vague; 0 = none.
- EX041 — Recommendations. List two recommendations. Rubric: 3 = account-level block, template fixes, fewer admins, private export; 2 = one; 1 = vague; 0 = none.
- EX042 — Figures. How are figures organised? Rubric: 3 = one per explanation, same scale; 2 = partial; 1 = vague; 0 = none.
- EX043 — Overclaiming. How was the overclaim fixed? Rubric: 3 = expected to reduce; test later; 2 = partial; 1 = vague; 0 = none.
- EX044 — Exposure vs access. Why is exposure not access? Rubric: 3 = data may not have been read; 2 = partial; 1 = vague; 0 = none.
Synthesis
Write clearly, limit claims and separate judgement from evidence.
Continue studying — chapter 07 of 08
← Chapter 06 (draft) · Course chapter index · Chapter 08 (draft) →