Selection bias distorts findings when some groups enter a sample more readily than others, making those groups overrepresented. It can turn a real-world question into a misleading answer. section 9.4
Everyday examples
Successful dieters are more likely to post before-and-after photos, so online posts can make a diet look more effective. An airline polling only its own passengers will probably hear favorable opinions. A workday phone poll may overrepresent retirees and stay-at-home parents while missing people at work. section 9.4
Before collecting or interpreting data, ask:
- Who is most likely to participate, and who is most likely to be missing? section 9.4
- Could the sampling time or location encourage particular groups to select themselves into the sample? section 9.4
- Can the population be defined around data that is already available, such as insured car owners or publicly traded companies? section 9.4
- Disclose the bias and plan mitigation steps; random sampling can reduce self-selection, though forcing participation is unethical. section 1.4 section 9.4
