under 7 different scenarios: ā1. Null scenario: no effect of treatment ā2. RCT scenario: treatment randomly assigned ā3. Confounder-adjusted ā4. Confounder-adjusted with instrument ā5. Confounder-adjusted and subject to selection bias ā6. Confounder-adjusted and subject to selection bias with instrument ā7. Confounder-adjusted and subject to selection bias with instrument and common cause of instrument and outcome With scenario number 7 being what we believe to be closest to the truth. |