A causal diagram sorts every candidate variable for the PHLT 8500 selection week into confounder, mediator or neither, and the adjustment set follows from the effect being estimated. Searches like "phlt 8500 week 3 assignment example", "phlt8500 week 3 sample" and "phlt 8500 week 3 example" land here.
What a finished PHLT 8500 Week 3 variable selection looks like
About three pages built around one directed acyclic graph. The graph places night shift work and incident hypertension at its ends and arranges age, sex, job category, BMI, sleep duration, smoking and years of service between and around them, with arrows drawn from stated evidence or reasoning. A classification table follows: each variable, its position, its role, and the decision. Age and job category sit upstream of both exposure and outcome and are adjusted for. Sleep duration and BMI sit downstream of shift work, on paths toward hypertension, and are treated as mediators, excluded when the question is the total effect. Smoking is argued both ways before a decision. A final section contrasts the diagram's adjustment set with what automated stepwise selection produced on the same data, and explains why the two differ.
How a PHLT 8500 Week 3 example is structured
The document separates the causal reasoning from the statistical output on purpose. The estimand comes first, the total effect of night shift work on hypertension, because confounder and mediator are roles defined by a target effect, not fixed properties of a variable. The diagram follows, then the classification table, then a paragraph for each contested variable. Only after the adjustment set is fixed does the document look at data, and then only to show what an automated procedure would have chosen instead. That comparison is the document's warning: stepwise selection rewards variables that predict hypertension, mediators included, and would have adjusted away part of the effect under study. Hosmer and Lemeshow's purposeful selection approach is named as a middle path that keeps subject-matter reasoning in charge. A closing note says what would change if the question were the direct effect.
The effect named before the variables
The total effect of night shift work on hypertension is stated as the target. Every later classification depends on that choice, and the document says which variables would change role if the direct effect were the goal instead.
Arrows with reasons
Each arrow in the diagram cites evidence or states the reasoning behind it. An arrow drawn because it seemed plausible is marked as an assumption, letting a reader separate the firm parts of the map from the judgment calls.
Upstream, adjusted
Age and job category influence both shift assignment and blood pressure. They sit on backdoor paths, and adjusting for them closes those paths, which is the definition of confounder control the document uses.
Downstream, left alone
Sleep duration and BMI plausibly change because of shift work and then affect pressure. Adjusting for them would remove part of the effect the question asks about, so they stay out of the total-effect model.
What stepwise would have done
An automated procedure run on the same candidates retained BMI and sleep duration because they predict hypertension well. The document shows that result and explains why predictive strength is the wrong criterion for this question.
Where marks go in PHLT 8500 Week 3
The reasoning behind each classification carries the analytic share. A document sorting variables into confounders and mediators without explaining the paths that justify each label has asserted a diagram rather than argued one, and graders mark that down. The estimand statement is checked first, since a classification with no stated target effect cannot be right or wrong. Mediators adjusted for in a total-effect model are the most costly error, because the specification then estimates something other than the effect posed. Contested variables earn credit when argued in both directions before a decision. The stepwise comparison is valued in doctoral sections for showing why prediction and causal adjustment diverge. Diagram clarity, labeled arrows and cited evidence for key paths account for the presentation share.
Get a PHLT 8500 Week 3 example written to your instructions
Give the desk the Week 3 prompt and rubric, your exposure and outcome, and the candidate variables. A selection document with a causal diagram, a role for each candidate and a reasoned adjustment set comes back inside 24-48h, and a first one is free. Its arrows describe a composite worker cohort; the map for your question must be argued from your own evidence.
PHLT 8500 Week 3 questions, answered
What is the difference between a confounder and a mediator?
A confounder influences both the exposure and the outcome and sits outside the causal path between them; adjusting for it removes a source of bias. A mediator lies on the path, changed by the exposure and in turn affecting the outcome; adjusting for it removes part of the effect itself. The same variable can play different roles depending on timing, which is why the diagram states when each was measured.
Is stepwise selection ever acceptable?
For pure prediction, some analysts still use it with caution, though it tends to overfit and to produce biased coefficients and overly narrow intervals. For estimating the effect of an exposure it is poorly suited, because it selects on association with the outcome and has no way to tell a confounder from a mediator. The document explains which purpose applies before judging the method.
Do I need special software to draw the causal diagram?
No. A clear figure in any drawing tool works, as long as arrows are unambiguous and each node is labeled. Free tools such as DAGitty can also identify adjustment sets from a drawn diagram, which some sections encourage as a check. The grade depends on each arrow being justified and on the adjustment set following logically from the picture.