Observational Studies vs. Experiments

Observational Studies

• Researchers collect data without interfering with how the data arises.
• Can only establish an association or correlation between explanatory and response variables.
• Retrospective Study: Uses data from the past.
• Prospective Study: Data collected throughout the study.

Experiments

• Researchers randomly assign subjects to various treatments.
• Can establish causal connections between the explanatory and response variables.

Random Assignment Example

• Observational Study: Sample individuals who choose to work out vs. those who don't; compare average energy levels.
• Experiment: Randomly assign individuals to work out or not; compare average energy levels.
• Key Difference: Decision to work out is imposed by the researcher in experiments, not subjects.
• Observational study may include uncontrolled variables that affect outcomes (e.g., being in better shape).
• Experiment controls for such variables through random assignment; allows for causal inference.

Media Coverage Study Example

• Study: Breakfast cereal keeps girls slim.
• Observational study tracked 2,370 girls from ages 9 to 19.
• Possible explanations for findings:
  1. Eating breakfast causes girls to be slimmer.
  2. Being slim causes girls to eat breakfast.
  3. A third variable (e.g., being health-conscious) causes both slimness and breakfast consumption.
• Confounding Variables: Affect both explanatory and response variable; can misleadingly imply a relationship.

Key Takeaway

• Correlation does not imply causation.
• Type of study (observational vs. experimental) determines whether we can infer causation.
• Observational studies generally only support correlational statements.
• Experiments can support causal conclusions.
• Advanced Methods: Causal inference methods can draw causal conclusions from observational studies, but this is beyond the scope of the course.