Multiple Bias Modeling

Real studies often have multiple biases:

1. Confounding — unmeasured variables
2. Selection bias — differential participation
3. Misclassification — measurement error

Combined effect: Biases may amplify or cancel each other.

Scenario	Result
All biases toward null	Severe underestimate
Biases in opposite directions	May partially cancel
Biases interact	Complex, non-additive effects