Multi-Category Original Variable

a) SAS/Stata Tasks.

i) Create six dummy variables to indicate persons of: 1) Hispanic ancestry and any race, 2) non-Hispanic ancestry and white only race, 3) non-Hispanic ancestry and African American only race, 4) non-Hispanic ancestry and American Indian or Alaskan Native only race (“AIAN”), 5)

non-Hispanic ancestry and Asian only race, and 6) non-Hispanic ancestry and multiple races. (Be careful to account for missing value codes) .

ii) Regress bmiR on all six dummy variables (we will refer to this as Regression #3 in the Write-Up Tasks).

iii) Regress bmiR on fi ve of the six dummy variables, using persons of nonHispanic white race-ethnicity as the reference category (we will refer to this as Regression #4).

(1) Use the test command to test whether there are any differences in the mean body mass index among the six race-ethnicity categories (we will refer to this as Test #1 in the Write-Up Tasks).

(2) Use the test command to test whether the mean body mass index differs between each pair of the fi ve categories included in the regression model (ten tests all together; we will refer to this as Test #2 in the Write-Up Tasks).

iv) Regress bmiR on fi ve of the six dummy variables, using persons of Hispanic race-ethnicity as the reference category (we will refer to this as Regression #5 in the Write-Up Tasks).

(1) Use the Stata lincom command to test whether the means differ between each pair of the fi ve categories included in the regression model (ten tests all together; we will refer to this as Test #3 in the Write-Up Tasks).

b) Write-Up Tasks.

i) Describe the results in Regression #3. Are they what you expected? How should you pick the reference category in a model with dummy variables?

ii) Write the null and alternative hypotheses and a conclusion based on the results of the Test #1 command. Is this test shown in the output of any of the regressions (Regression #3, Regression #4, Regression #5)? If it is, why? If not, why not?

iii) Interpret the point estimates of the intercept and dummy variable coeffi cients in the default output for Regression #4.

iv) Interpret the point estimates of the intercept and dummy variable coeffi cients in the default output for Regression #5.

v) Indicate which of the tests in Test #2 and Test #3 correspond to which of the tests in the default output of Regression #4 and Regression #5. Be sure to note whether any of the tests in Test #2 and Test #3 are not shown in either set of default output.