The BRFSS linear regression model of Section 5.12 was fit accounting for the sampling design and weights (using PROC SURVEYREG), with relevant output shown here.
Answer the same set of questions previously addressed in Section 5.12 using the output for this new weighted analysis. The degrees of freedom are different for this analysis (df = 420,993) and that relevant standard errors are provided, since they are more challenging to compute manually.
a. What is the form of the calculated straight-line fit using alcohol frequency ("drink days") as a predictor of BMI?
b. Determine if there is a statistically significant linear relationship between alcohol frequency and BMI.
c. What is the estimated mean BMI and corresponding 95% confidence interval for a woman who drinks the median number of 4 days per month? For one who drinks 15 days per month (90th percentile of consumption frequency)?
d. Comment on how these results do, or do not, differ from those for the unweighted analysis.