Reading

Background material on correlation and causation

Science for Business, Law and Journalism. 2000. Read the following material: Chapter 13. Holes in the fabric of evaluation Chapter 14. Correlations are hard to interpret, Chapter 16. Controls. Note: These three chapters provide basic background on correlations, controls and causation.

Apply these ideas to Grandjean et al. (1998)

Grandjean et al. 1998. Find examples in Grandjean et al. of the major concepts in the Science for Business, Law and Journalism chapters cited immediately above.

Questions for class discussion

1. Your instructor will ask you for a simple sentence that correctly uses some or all of the following words: correlation, control for, control group and confounding. Be prepared to provide a sentence that correctly uses each of these concepts, using the facts in Grandjean et al. (1998).

2. Carefully read the Abstract of Grandjean et al. (1998) What correlations did the author's find?

3. Give an example of a factor that the Grandjean et al. study controls for, and a factor that it does not control for.

4. What confounding variables are Grandjean et al. concerned with?

5. What exactly does "correlation does not imply causation" mean? Does it mean that if two variables are correlated, they can never be related causally?

6. Compare the scientific view of causation presented above to the legal "but for" definition of causation, (as used in torts, for example). Are there circumstances under which science could prove causation, but a tort plaintiff would definitely not meet the burden of proving causation? Is the reverse situation possible?

7. Explain how randomization helps avoid confounding. Explain the key distinction between an experiment and an observation. Explain the benefits and drawbacks of each. Is is possible for an observational study to be randomized in a useful way? An experimental study?