Part A of Computational Public Health Statistics Final Exam / S99

Instructions: Please write your answers on the separate page provided. Please number your responses. You may do this part of the exam either before or after Part B, but once you turn it in, there is no going back to it. This part of the exam should take no more 30 minutes to complete. [Square brackets contain point values.]

1. Using statistical notation, state the null hypothesis tested by case-control studies. [1]

2. You are performing a test to compare an observed proportion to a hypothetical "null" (p₀). Where might p₀ come from? [2]

3. You are performing an ANOVA comparing three independent groups. Write the null and alternative hypotheses that are being tested. [2]

4. The Kruskal-Wallis is said to be more robust than ANOVA because it can be used under a greater number of conditions. Why is this so? [2]

5. When is Fisher's test used in place of a chi-squared test? [1]

6. What distinguishes experimental studies from observational studies. [1]

7. What distinguishes case-control studies from cohort studies? [1]

8. Define confounding. [2]

9. Define interaction. [2]

10. What are the necessary preconditions for confounding? [2]

11. We find significantly different odds ratios in different strata. What does this suggest? [1]

12. An adjusted (Mantel-Haenszel summary) relative risk leaves us with a distinctly different impression of risk compared to a crude relative risk. What does this suggest? [1]