Evaluating Causal Hypotheses

Introduction

• Previous lecture discussed the concept of causation:
  • Effect on an individual (deterministic or probabilistic)
  • Effect on a population
• Difficulty in determining an individual's complete residual state
• Use of population to determine causal effects

Focus of the Lecture

• Evaluating causal hypotheses
• Focus on randomized experimental trials:
  • Considered the gold standard
• Next lecture: Other experimental designs

Course Announcements

• Homework for Chapter 7 deadline extended
• Exam format change:
  • Take-home exam
  • Open book
  • Focus on understanding concepts rather than precise details

Example Study: Randomized Experimental Trial

• 1977 FDA study on saccharin and rats
• Study showed saccharin causes cancer in lab rats
• FDA issued a warning label rather than a ban

Study Details

• Two-generation rat feeding experiments:
  • Both generations received saccharin
  • Control group did not ingest saccharin
• Significant evidence found in the second generation
• First generation results were not statistically significant

Experimental Design

• Causal variable: Ingestion of saccharin
• Effect variable: Occurrence of bladder cancer
• Randomized sampling and division into control and experimental groups

Evaluation of Causal Hypothesis

• Determining statistical significance:
  • Use of P values
  • Significant if P < 0.05
• Inference to humans from rat studies
• Threshold effect considerations

Methodology

1. Identify real-world population and causal hypothesis
2. Analyze data
3. Identify experimental design
4. Evaluate random sampling
5. Evaluate hypothesis
6. Provide summary

Considerations for Human Application

• Large doses of saccharin used in rats
• Human equivalent: 800 bottles of soda per day
• Risk assessment and public policy implications

Double-Blind Trials

• Ensures only systematic difference is the causal variable
• Placebo effect and psychological impacts
• Double-blind: Neither subjects nor experimenters know group assignments

Conclusion

• Importance of randomized experimental trials in evaluating causal hypotheses
• Homework and office hour reminders