A. A study was conducted to examine the relation between body weight and blood pressure among adults 60 years of age and older. The hypothesis is that higher body weight will be associated with higher values of blood pressure while lower body weight will be associated with lower blood pressure values.

1. It is reasonable to use a logistic regression to examine the relation between body weight and blood pressure
o True
o False
2. If blood pressure were normally distributed, then plotting blood pressure residuals on a X-Y axis would yield a line with a negative slope
o True
o False
3. The study reported a regression coefficient (beta coefficient) of +0.3 and an intercept of 135 from a linear regression using blood pressure as a dependent variable and body weight as exposure.
Write the equation predicting blood pressure (mm of mercury) as a function of body weight (kilograms).

__________________________________________________________________________

4. Explain in plain English what the regression coefficient and the slope of the above equation mean
___________________________________________________________________________
__________________________________________________________________________

5. After adjustment for sex and exercise, the regression coefficient for body weight became 0.2. Is there confounding by sex and exercise? Justify your answer.

__________________________________________________________________________

6. What is the predicted change in blood pressure if a women weighing 100 kg loses 10 kg of weight after treatment? Answer this question using the most valid equation.

_________________________________________________________________________

B. A second study investigates the relation of smoking with prevalence of diabetes in 5000 adult men and women using employment data. Unfortunately, it was not possible to follow up these employees over time. Hence, investigators are relying on cross-sectional data.

7. What type of regression will you use to answer the study question? ____________________

8. Write the crude equation assuming that the intercept is 0.0005 and the beta coefficient for smoking is 0.6.

___________________________________________________________________________

9. What is the probability that smoker has diabetes (smoking=1)?

__________________________________________________________________________

10. What is the probability that a non-smoker has diabetes (smoking=0)?

_________________________________________________________

C. The crude incidence rate of prostate cancer in a US male population is 8 cases per 1000 person years. You wish to compare that rate observed in the US with the rate of prostate cancer among Australian male residents. To address potential confounding by age, you create 3 age groups (<40, 40-59, and 60+ year). The age distribution of men in the US for those three age groups is 25%, 30%, and 45%, respectively. The age-specific incidence rate for prostate cancer in the US is 3/1000 person-years for men <40 years, 7/1000 person-years for men aged 40-59, and 9/1000 person-years for men 60+ years of age.

11. Using the total US male population as standard, calculate the age-standardized incidence rate of prostate cancer among US males.
_______________________________________________________________________

12. Is age a confounding factor? Justify your answer.

________________________________________________________________________

13. Assuming that the Australian population is much younger than the US male population, can we compare the crude incidence rates of prostate cancer between the 2 countries? Justify your answer.

___________________________________