| Course | PHLT 8520 Advanced Epidemiological Methods |
|---|---|
| Week | Week 1 |
| Paper type | Error taxonomy brief |
| Length | About 690 words, 5 pages |
| Format | APA 7 student paper |
| School | Walden University |
| Program | PhD in Public Health |
| Updated | October 2026 |
Free sample paper for PHLT 8520 Week 1
Would a Bigger Study Shrink It? An Error Taxonomy for a Composite Cohort Asking Whether Night Shifts in Early Pregnancy Raise the Risk of Preterm Birth
Student Name
Walden University
PHLT 8520: Advanced Epidemiological Methods
Instructor Name
Month Day, Year
Would a Bigger Study Shrink It? An Error Taxonomy for a Composite Cohort Asking Whether Night Shifts in Early Pregnancy Raise the Risk of Preterm Birth
The Question the Taxonomy Serves
Every example below comes from one composite study question: among employees of a large health system, does working at least one night shift between gestational weeks 4 and 20 raise the risk of delivering before 37 completed weeks? A night shift follows the definition used in Danish payroll research, at least three working hours between 11 p.m. and 6 a.m. (Specht et al., 2019). The taxonomy sorts the ways an estimate from such a study could depart from the truth, so that later weeks can ask how far and in which direction.
One Sorting Rule
An error is random if doubling the sample would shrink it. It is systematic if a larger study would only report the same distortion with a narrower interval. Rothman and Greenland (2005) draw the same line between precision and validity, and it decides the first branch before any named bias enters the page. Labels such as recall bias or Berkson bias are not branches here; each is filed under the structure that produces it.
The Branching Diagram in Words
Table 1
Error Taxonomy for the Composite Night-Shift and Preterm Birth Cohort
| Branch | Structure | Composite example | Would a larger sample shrink it? | |
|---|---|---|---|
| Random error | Sampling variation | A pilot of 400 pregnancies yields 9 preterm births among night workers; one more or one fewer moves the ratio noticeably | Yes | |
| Systematic: confounding | A shared cause of exposure and outcome | Lower job grade raises the chance of night rosters and, through income and job strain, the chance of preterm delivery | No | |
| Systematic: selection | Entry or retention depends on factors tied to exposure and outcome | Women whose early complications took them off night rosters before enrollment are counted as day workers or lost | No | |
| Systematic: measurement | Exposure or outcome recorded with error | Night work read from job titles rather than timekeeping, so a staff nurse who never worked nights is coded as exposed | No | |
Note. Composite examples written for this sample; no real cohort is described.
Chance in a Small Cohort
Random error is easiest to see in a small study. With nine preterm births among a few hundred night workers, the estimate could swing from a modest excess to a modest deficit as single deliveries move, even with a flawless design. A confidence interval describes that swing. It says nothing about the three systematic branches, which is why a narrow interval in a large biased study can mislead more than a wide one in a small clean study.
Three Structures That Survive a Larger Sample
Confounding is drawn as a fork: one variable points to both night work and preterm birth. Selection is drawn as a gate: who enters, or stays, depends on things connected to both, a structure Hernán et al. (2004) describe as conditioning on a common effect. Measurement error is drawn as a faulty copy: the variable analyzed differs from the variable that matters. Each structure has a remedy of its own, and naming the structure first is what tells the analyst which remedy to reach for.
One Label, Two Branches
The healthy worker effect shows why labels make poor branches. At hire, women with chronic conditions may avoid night rotations, so night workers start out healthier; that is confounding by baseline health. After conception, women who develop early bleeding or hypertension may be moved to day work or leave, so the night group is purged of its highest-risk pregnancies; that is selection, and Arrighi and Hertz-Picciotto (1994) call the second process the healthy worker survivor effect. Treating both as one bias invites a single remedy, adjustment for baseline health, that fixes only the first.
What the Taxonomy Postpones
Sorting is only the first judgment. The taxonomy does not say which way each error pushes the estimate or by how much, and it does not rank them. Those questions belong to the weeks on confounding, selection, measurement and sensitivity analysis, where each branch is examined with the composite cohort's own data.
References
Arrighi, H. M., & Hertz-Picciotto, I. (1994). The evolving concept of the healthy worker survivor effect. Epidemiology, 5(2), 189-196. https://doi.org/10.1097/00001648-199403000-00009
Hernán, M. A., Hernández-Díaz, S., & Robins, J. M. (2004). A structural approach to selection bias. Epidemiology, 15(5), 615-625. https://doi.org/10.1097/01.ede.0000135174.63482.43
Rothman, K. J., & Greenland, S. (2005). Causation and causal inference in epidemiology. American Journal of Public Health, 95(S1), S144-S150. https://doi.org/10.2105/AJPH.2004.059204
Specht, I. O., Hammer, P. E. C., Flachs, E. M., Begtrup, L. M., Larsen, A. D., Hougaard, K. S., Hansen, J., Hansen, Å. M., Kolstad, H. A., Rugulies, R., Garde, A. H., & Bonde, J. P. (2019). Night work during pregnancy and preterm birth: A large register-based cohort study. PLOS ONE, 14(4), Article e0215748. https://doi.org/10.1371/journal.pone.0215748
PHLT 8520 Week 1 instructions, in plain terms
Opening assignments in advanced epidemiology tend to ask for an organized account of error before any analysis is attempted. Expect a prompt that names random error, confounding, selection and information bias, and asks you to show how they differ with examples from a single study question you carry through the term. Instructors look for a classification principle rather than a glossary. A table or branching diagram is common, and two pages of text is a typical length. Some sections let you pick your own exposure and outcome; others supply one. Citing at least one methods text or article on bias structure is usually required. A short note on what the taxonomy does not yet decide, such as direction and size, is often welcomed because it shows you see where the term is heading.
Inside the PHLT 8520 Week 1 example
Built as a two-page brief, the sample first fixes the study question, the exposure window and the payroll definition of a night shift. It then states the sorting rule and credits it to the distinction between precision and validity. Table 1 carries four rows, one random and three systematic, each with its structure, a composite example and a yes or no answer to the sorting question. Short sections follow on chance in a small cohort, on the three systematic shapes drawn as a fork, a gate and a faulty copy, and on the healthy worker effect straddling two branches. Two margin notes mark the moves a grader is likely to reward. A last paragraph lists what the taxonomy postpones to later weeks. Four checkable sources, including a Danish payroll cohort, support the claims.
Where the marks sit in the PHLT 8520 Week 1 rubric
Rubrics for this first brief weigh classification logic most heavily. The sample collects that credit by stating one test, whether more data would shrink an error, and applying it to every entry. Filing bias by structure, rather than listing named biases, earns a further share because it proves the author reasons about mechanisms. Examples drawn from a single composite cohort show the taxonomy can be used, not just recited. The healthy worker section is the strongest item, since splitting one label into confounding and selection shows the categories were tested against a hard case. Clear writing, a readable table and correctly formatted citations take the smaller rows. Instructors also note whether random error is ever described as a kind of bias; this brief keeps the two apart throughout.
Common PHLT 8520 Week 1 mistakes, and how to avoid them
Error taxonomies lose marks when they become lists of thirty named biases with one-line definitions. Choose a sorting rule first and let it decide where each error goes. Keep random error out of the bias branch, and avoid suggesting that a small p-value rules out chance error in a design. Use one study question for every example so the reader sees each error in a real setting. Test your categories on a label that resists them, such as the healthy worker effect or recall bias. Say what the taxonomy cannot yet tell you. If your instructor assigns a different exposure, the same frame still works; send us your question and a first custom brief built around it is free.
Get a PHLT 8520 Week 1 example written to your instructions
Send the Week 1 prompt and the rubric, and mention the study question your section is using, if one was set. Your taxonomy comes back with its sorting rule stated first, each distortion filed by what produces it, and examples drawn from that question. The first is free and returns within 24-48 hours.
More PHLT 8520 papers and PhD in Public Health samples
- PHLT 8520 Week 2: Design Selection Paper
- PHLT 8520 Week 3: Confounding Brief
- PHLT 8520 Week 4: Design Proposal
- PHLT 8520 Week 5: Selection Analysis
- PHLT 8520 Week 6: Peer Challenge Reply
- PHLT 8520 Week 7: Measurement Error Brief
- PHLT 8520 Week 8: Control Strategy Comparison
- PHLT 8520 Week 9: Sensitivity Argument
- PHLT 8520 Week 10: Validation Check
- PHLT 8520 Week 11: Final Methods Paper
- PHLT 8500 Week 1: Regression Review Post
- PHLT 8450 Week 1: Community Definition Post
- PHLT 8331 Week 1: Population and Sample Concept Post
- PHLT 8270 Week 1: Surveillance Coverage Post
PHLT 8520 Week 1 questions, answered
Where can I find a free PHLT 8520 Week 1 sample paper?
A full error taxonomy brief for Week 1 is posted on this page, with its sorting rule, a four-row table, margin notes and real references on bias structure.
What is the difference between random and systematic error in epidemiology?
Random error shrinks as a study grows, because it comes from sampling variation. Systematic error stays the same however large the study, since it comes from how the study was designed, who entered it or how variables were measured.
Is the healthy worker effect confounding or selection bias?
It can be both. Healthier women choosing night work at hire behaves like confounding, while women leaving night work after early complications behaves like selection, which is why the sample files it on two branches.
Why sort biases by structure instead of by name?
Names multiply and overlap, but structures are few. Knowing whether a bias is a shared cause, a condition on entry or a measurement fault tells you which remedy to use.
Is the night-shift cohort in this sample real?
No. The cohort and its figures are composite, written for this example. The night-shift definition and the background evidence come from a real Danish payroll study, cited in the references.