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 Robustness"]
N3["Lab — Leave-one-month-out check on the level change"]
N4["7.3 Useful limitations"]
N5["7.4 Implications"]
N6["Synthesis"]
N0 --> N1 --> N2 --> N3 --> N4 --> N5 --> N6
N6 --> EX["Worked example"]
EX --> Q["Exercise EX039"]
Logical drawing
flowchart LR
N0["Learning outcomes"]
N1["7.1 Structure"]
N2["7.2 Robustness"]
N3["Lab — Leave-one-month-out check on the level change"]
N4["7.3 Useful limitations"]
N5["7.4 Implications"]
N6["Synthesis"]
N0 --> N1 --> N2 --> N3 --> N4 --> N5 --> N6 --> Q["EX039"]
[Learning outcomes]
|
v
[7.1 Structure]
|
v
[7.2 Robustness]
|
v
[Lab — Leave-one-month-out check on the level change]
|
v
[7.3 Useful limitations]
|
v
[7.4 Implications]
|
v
[Synthesis]
|
v
[Worked example] --> [Exercise EX039]
Learning outcomes
You will (1) structure a dissertation chapter by chapter; (2) write limitations that inform rather than excuse; (3) test the robustness of the main result.
7.1 Structure
Sofia's dissertation follows a clear structure: introduction, literature and theory, methods, results for each part, integrated discussion, limitations, implications and conclusion. Each results chapter opens with the question it answers and ends with a short summary. She writes the methods chapter first, while details are fresh, and the introduction last, once she knows what she found. Dr Nakamura reads chapters as they are drafted and asks her to cut repetition between the literature review and the discussion. She keeps a running list of decisions to explain in the methods chapter.
7.2 Robustness
Before writing the conclusion, Sofia checks whether her main result depends on choices she made. She reruns the time series analysis dropping each month in turn, to see whether any single unusual month drives the result. The lab shows the level change stays within a narrow range whichever month is dropped. She also tries counting resets per staff member rather than per account, with similar results. Robustness checks show readers that the conclusion would not collapse if a reasonable analytic choice were made differently.
Lab — Leave-one-month-out check on the level change
# Leave-one-month-out check on the interrupted time series level change (same synthetic data as lab_its).
import random
random.seed(3)
y = [62 - 0.2 * t + random.gauss(0, 2) for t in range(18)] + [39 - 0.25 * (t - 18) + random.gauss(0, 2) for t in range(18, 30)]
def fit(xs, ys):
n = len(xs); mx = sum(xs) / n; my = sum(ys) / n
b = sum((x - mx) * (v - my) for x, v in zip(xs, ys)) / sum((x - mx) ** 2 for x in xs)
return my - b * mx, b
def level(skip):
pts = [(t, v) for t, v in enumerate(y) if t != skip]
pre = [p for p in pts if p[0] < 18]; post = [p for p in pts if p[0] >= 18]
a0, b0 = fit([p[0] for p in pre], [p[1] for p in pre]); a1, b1 = fit([p[0] for p in post], [p[1] for p in post])
return (a1 + b1 * 18) - (a0 + b0 * 18)
vals = [level(s) for s in range(30)]
print(f"full data level change {level(-1):+.1f}; leave-one-out range {min(vals):+.1f} to {max(vals):+.1f}")
Run with python3 lab_robust.py. Verified output:
full data level change -18.8; leave-one-out range -19.3 to -17.8
7.3 Useful limitations
Sofia writes limitations that tell readers how to interpret the work. The study covers one trust, so results may differ in organisations with different staff or systems. Survey groups are not matched individuals. The password manager's effect cannot be separated from training gaps. The study did not measure security outcomes such as account compromise, because those data were not available. For each limitation she states its likely direction of bias and what kind of study could address it. She keeps the section short and specific to avoid diluting it.
7.4 Implications
Sofia separates implications for the trust, for other hospitals and for policy. For the trust: keep the change, provide manager training and track compromise indicators. For other hospitals: the change reduced burden here and is worth trialling with measurement. For policy: guidance against routine forced changes is consistent with these findings, though one study cannot settle the question. Writing implications at three levels prevents the overreach that occurs when single-site findings are presented as universal truths. Each implication cites the specific result that supports it, so that readers can trace every recommendation back to evidence.
Worked example
Sofia structures the dissertation, checks robustness by leaving out each month, writes directional limitations and states implications at three levels.
Synthetic case. Results are synthetic.
Errors and limits
Robustness checks cover only chosen alternatives. Implications may be read more broadly than intended.
Exercises
- EX039 — Writing order. Which chapters does Sofia write first and last? Rubric: 3 = methods first, introduction last;; 2 = order;; 1 = vague;; 0 = none.
- EX040 — Leave-one-out. What does dropping each month in turn test? Rubric: 3 = whether one unusual month drives the result;; 2 = influence;; 1 = vague;; 0 = none.
- EX041 — Robust range. What range did the leave-one-month-out lab show? Rubric: 3 = level change between -19.3 and -17.8;; 2 = narrow;; 1 = wide;; 0 = none.
- EX042 — Missing outcome. Which security outcome was not measured? Rubric: 3 = account compromise;; 2 = compromise;; 1 = resets;; 0 = none.
- EX043 — Trust implications. What did Sofia recommend to the trust? Rubric: 3 = keep the change, train on the manager and track compromise;; 2 = three;; 1 = vague;; 0 = none.
- EX044 — Three levels. At which three levels are implications stated? Rubric: 3 = trust, other hospitals and policy;; 2 = two;; 1 = vague;; 0 = none.
Synthesis
Robustness checks and directional limitations make a dissertation's conclusions credible and correctly scoped.
Continue studying — chapter 07 of 08
← Chapter 06 (draft) · Course chapter index · Chapter 08 (draft) →