What you'll learn
How to gather data so conclusions are valid (~12–15%).
Sampling
- Simple random sample (every group equally likely), stratified (sample within groups), cluster, systematic.
- Bias to avoid: voluntary response, convenience, undercoverage, nonresponse, response bias. Good sampling allows generalizing to the population.
Observational study vs experiment
- Observational: measure without imposing treatment → can show association, not causation (lurking variables).
- Experiment: impose treatments → can establish causation.
Experimental design principles
- Control (compare to a baseline/placebo), randomization (random assignment creates comparable groups), replication (enough subjects).
- Blinding (single/double) reduces bias.
- Blocking: group similar units, then randomize within blocks (controls a known variable); matched pairs is a special case.
Exam tips
- Random sampling → generalizable; random assignment → causal.
- Name and explain the design principle in context.
Common mistakes
- Confusing random sampling with random assignment.
- Claiming causation from an observational study.