A vaping study has causal arrows: read the assumptions behind them
Map exposure, outcome and selection assumptions in a causal DAG without treating an arrow as a discovered effect.
No Vape · Updated

Read the legend before interpreting the drawing
Locate what each node represents and whether the figure distinguishes measured variables, unmeasured variables or selection into the analysis. The same short label can mean a baseline condition in one report and a later observation in another. Keep those times attached to the labels. An arrow running between two words is not a measured numerical effect.
IARC’s original chapter on causal diagrams explains that DAGs encode hypothesized causal structure and can make bias assumptions explicit. Their usefulness depends on those assumptions. We use the chapter’s introductory distinctions without applying a formal bias judgment to a named vaping study or selecting adjustment variables for an analysis.
International Agency for Research on Cancer: Chapter2: Causal diagrams to evaluate sources of bias
Follow one stated pathway rather than every line at once
First mark the exposure and outcome named by the actual research question. Then trace a route through the figure and describe the proposed role of the intermediate nodes. A potential common cause, an intermediate mechanism and a variable affected by two other variables occupy different positions. Familiar labels do not fix those roles across all questions.
For an original fictional sketch, suppose a study proposes that two separate characteristics influence whether an adult joins an online study. Restricting the analysis to joiners selects on a shared consequence. This illustrates the question raised by a collider structure; it does not establish that selection biased a real quitting result or predict the direction or size of any bias.
Ask what was conditioned on and why
Conditioning may appear as statistical adjustment, restriction, stratification or selection. Locate the authors’ explanation rather than assuming the adjusted variable list is always a list of protective corrections. IARC explains why conditioning on a collider can introduce a noncausal association under the represented structure. The graph’s complete pathways still matter.
Do not make a rule that a certain variable should always be included or always omitted. Its role depends on the causal question and assumed relationships. A graph can also reveal an unmeasured assumption that the data alone cannot settle. Technical choices about sufficient adjustment sets need qualified methodological interpretation, not a home calculation based on one arrow.
Describe the assumption instead of announcing a causal verdict
Write a reading note with the question, the key proposed pathway, the relevant selection or adjustment choice and an unresolved assumption. If two graphs differ, ask what underlying relationship differs. Do not choose the graph that gives the most attractive quitting result. A disagreement about an arrow may be substantive even when both diagrams look polished.
This task differs from merely recognizing an adjusted estimate: it explains the structural argument behind a choice. It also differs from proving a mechanism. A personal trigger map is an ordinary planning aid and does not establish a causal DAG. For treatment decisions, seek suitable professional interpretation of the full evidence rather than turning a figure into a personal prescription.
Sources and scope
Sources inform the guidance. The worksheets and examples are No Vape editorial suggestions, not tested treatments.
Written with AI assistance and checked against the cited sources. No clinical review is claimed. This guide covers practical planning, not individual medical treatment. How we research and write