A third variable is associated with both the exposure and the outcome, creating a spurious or distorted association. Proper control (e.g., stratification or regression adjustment) brings the estimate toward the truth.
The true effect of the exposure differs across levels of another variable (e.g., stronger in smokers than nonsmokers). This is not a bias; we should report stratum‑specific effects or include an interaction term.
Stratification splits data by levels of a variable to examine subgroup effects. Adjustment controls for variables (e.g., via regression or Mantel–Haenszel) to estimate the association as if groups were similar on those variables.
A confounder is linked to both exposure and outcome, creating a backdoor path that distorts the crude estimate.
An effect modifier changes the strength/direction of the exposure → outcome effect (a true interaction). The arrow points to the relationship, not directly to outcome.
Enter cell counts to compute crude OR/RR and 95% CI. Use the scenario editor above for stratified (Mantel–Haenszel) estimates.
| Disease (+) | Disease (−) | |
|---|---|---|
| Exposed | ||
| Unexposed |
SE = √(1/a+1/b+1/c+1/d) on the log scale.Crude OR = 2.0 (significant). After adjusting for age, OR = 1.05 (NS). What is age here?
RR of treatment on outcome is 0.6 in men and 1.0 in women. Best description?
Scenarios below are didactic and mirror patterns described in epidemiology teaching (coffee–smoking–lung cancer; aspirin–sex–MI; asbestos–smoking–lung cancer).