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 Linking literature"]
N3["7.3 Limitations"]
N4["7.4 Practice implications"]
N5["7.5 Writing style"]
N6["Presenting results"]
N7["Contribution"]
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 Linking literature"]
N3["7.3 Limitations"]
N4["7.4 Practice implications"]
N5["7.5 Writing style"]
N6["Presenting results"]
N7["Contribution"]
N0 --> N1 --> N2 --> N3 --> N4 --> N5 --> N6 --> N7 --> Q["EX039"]
[Learning outcomes]
|
v
[7.1 Structure]
|
v
[7.2 Linking literature]
|
v
[7.3 Limitations]
|
v
[7.4 Practice implications]
|
v
[7.5 Writing style]
|
v
[Presenting results]
|
v
[Contribution]
|
v
[Worked example] --> [Exercise EX039]
Learning outcomes
You will (1) write findings clearly; (2) state limitations honestly; (3) connect to literature and practice.
7.1 Structure
The dissertation follows question, literature, design, results and discussion, with each figure traceable to code.
Traceability means each table and figure is produced by a named script from the extract, and the dissertation cites the script beside the figure. The second examiner relies on this when re-running the analysis. Chapter order follows the logic of the study: the reader meets the three explanations and their predictions before seeing any results, so the results can be read as tests. The appendices hold the dated analysis plan, the pilot-driven rules for missing fields and the dated predictions Chloé added on Dr Mehta's advice, so a reader can check that each result was tested as planned.
7.2 Linking literature
The admin finding supports least privilege (Saltzer and Schroeder, 1975) and CIS Controls v8 safeguards on administrative privileges; the operator perspective in Dietrich et al. (2018) helps explain why admin rights were granted widely.
Dietrich et al. matter here because they explain a pattern Chloé's data cannot: why Halvorsen granted standing admin rights to 120 users in the first place. Operators under time pressure grant broad rights to reduce support calls. Linking the two lets the discussion recommend a process, temporary elevation on request, rather than simply blaming users.
7.3 Limitations
One organisation; six months; observational design; monthly checks miss changes reverted within a month.
Each limit has a direction. One organisation means results may not hold for fleets with different management. Six months may be too short to see seasonal patterns. The observational design shows association, not cause. And a setting changed and restored between monthly checks is never counted, so drift is probably underestimated, most likely among admin devices, which can change and restore settings quickly. None of these limits reverses the main finding, but each tells the reader what the follow-up study should do differently.
7.4 Practice implications
Halvorsen can reduce standing admin rights and should re-measure drift after six months.
Removing standing admin rights is the change with the strongest support, and the follow-up gives Halvorsen a way to check it. If drift among the former admin devices falls towards the 15.0% seen among non-admin devices, the association looks causal; if it barely moves, other explanations need another look. Halvorsen's same-day elevation process is part of the change, because removing rights without a quick alternative pushes users to work around the baseline.
7.5 Writing style
Chloé reports counts with percentages and avoids causal language.
Every percentage in the dissertation comes with its count, such as 66 of 120 rather than 55% alone, so a reader can see how many devices each figure rests on. When Dr Mehta found a sentence saying admin rights "cause" drift, Chloé changed it to "are strongly associated with", consistent with the rest of the dissertation and with her observational design. The chi-square values appear with the 3.84 threshold, so readers can see at once why engineering is described as unclear rather than weak.
Presenting results
Chloé presents results in a table with groups, counts, percentages, risk ratios and test statistics, followed by a short paragraph interpreting each. She adds a figure showing drift by setting for admin and non-admin devices, which makes the pattern easier to see.
Each table has a note stating the data period and that devices are pseudonymised.
Contribution
The dissertation contributes a measured comparison of three common explanations at one organisation and a reusable analysis plan for others.
Reader testing
She asks Sven to read the discussion and confirm that practical implications are clear for an IT manager.
Word budget
She keeps the discussion within the word limit by cutting background that does not support the argument.
Ethics statement
The dissertation includes an ethics statement describing approvals, pseudonymisation, department-level reporting and data deletion, so readers can judge how participants were protected.
Final checks
Before submission, Chloé re-runs all code from a clean copy and confirms every number in the text matches the outputs. She also asks a classmate to read the limitations section for clarity.
Worked example
Dr Mehta notices a sentence saying admin rights "cause" drift. Chloé changes it to "are strongly associated with".
Synthetic case. Synthetic.
Errors and limits
Limitations written last are often weak; draft them alongside results.
Exercises
- EX039 — Literature link. Which sources support the admin finding? Rubric: 3 = Saltzer-Schroeder, CIS Controls v8; 2 = partial; 1 = vague; 0 = none.
- EX040 — Dietrich role. How does Dietrich et al. (2018) help? Rubric: 3 = explains why admin rights were granted; 2 = partial; 1 = vague; 0 = none.
- EX041 — Limitations. List the four limitations. Rubric: 3 = one org, six months, observational, monthly checks; 2 = three; 1 = two; 0 = none.
- EX042 — Monthly gap. What do monthly checks miss? Rubric: 3 = changes reverted within a month; 2 = partial; 1 = vague; 0 = none.
- EX043 — Wording fix. How was the causal sentence fixed? Rubric: 3 = strongly associated with; 2 = partial; 1 = vague; 0 = none.
- EX044 — Next step. What should Halvorsen do next? Rubric: 3 = reduce admin, re-measure; 2 = partial; 1 = vague; 0 = none.
Synthesis
Write clearly, link to literature and state limits honestly.
Continue studying — chapter 07 of 08
← Chapter 06 (draft) · Course chapter index · Chapter 08 (draft) →