Causal reasoning · 4/6
Build a comparison: what if nothing changed?
Separate before–after change from an effect and state comparison assumptions.
About 15–20 minutes. The following cases are fictional.
Record your first judgment
A shop changes its sign: weekly sales rise from 100 to 130. A similar unchanged shop rises from 80 to 100. The owner attributes all 30 extra items to the sign.
Understand the judgment process
Last week is not a complete substitute for this week without the sign. Weather, holidays and demand can change. A causal contrast concerns outcomes under two intervention states, only one of which is observed for the same unit and period.
The difference in changes is (130−100)−(100−80)=10 items. This descriptive calculation becomes a causal interpretation only under assumptions: without the sign, changes would have been comparable; no differential promotion, customer spillover or measurement change explains the gap.
Examine multiple pre-intervention periods and business context. Similar past trends support but do not establish parallel trends. One comparison and one period also leave uncertainty about stability.
Checks you can perform
- Define the intervention date and common outcome units.
- Find comparable controls unaffected by the intervention.
- Compare changes and inspect simultaneous events.
- Separate arithmetic, identification assumptions and uncertainty.
Guided practice
Treatment rises 30 and comparison rises 20. Calculate the difference and state a conditional conclusion.
Explore the explanation
Ten items. Interpretation as an effect requires comparable untreated trends, no differential concurrent events and no spillover, among other conditions. The numbers cannot verify these assumptions.
Apply it in a new context
A school changes teaching methods and scores rise eight points, but the exam is easier. What comparison would help?
Compare after attempting
Use comparable classes taking the same exam at the same time without the method change. Examine prior trends and composition. Raw scores from different-difficulty exams are not a common outcome.
Preserve a revision record
Differences in changes do not remove every bias. A control affected by the intervention may invalidate the comparison.
Check whether you supplied inspectable evidence, a feasible next step and revision conditions. Revealing an answer is not mastery.
Source and scope
Hernán & Robins — Causal Inference: What If (chapters 1–3, 7–8)
The source provides conceptual or methodological background. Cases, steps and exercises are authored here; they are not the original experiment or evidence of this course’s effectiveness.
Completion is a personal record, not proof of mastery or certification.