PA 705
Fall 2004

Homework Assignment # 7

Regression Analysis and Correlation

Due Nov. 17

I. Please do ORB problem #2 (p. 456) and #7 (p. 458).

II. Lab Work

Please attach your output.

Open Employee.sav

1. Choose two variables that you expect to be positively or negatively correlated. State your hypothesis and run the procedure to test it.

1. Was your hypothesis supported?
    

2. How strong is the relationship?

H₁ As education increases, so does salary.

H_o There is no relationship between education and salary.

2. Test this hypothesis with a scatterplot. What does it suggest?
    

3. Test the hypothesis with a correlation procedure. What does your finding suggest?
    

4. Run a bivariate regression procedure.

   1. Is the hypothesis supported?
       

   2. What would be your best guess as to what the salary would be for an employee with no education?
       

   3. For each year of education, how much salary is earned?
       

5. After doing additional research, we find that salary is often related to one's seniority in the agency. Add this second independent variable (months since hire).

   1. How much, if any, does seniority add to the explanatory power of our model?
       

   2. Does education or seniority have a greater impact on salary? What makes you say this?
       

6. We then find, to our dismay, that there is an error in recording months since hire and so the variable is not reliable.  Meanwhile, our ongoing research has suggested that women and people of color are often paid less than men and nonminorities.

   1. Re-run the regression procedure. Delete months since hire and including gender and race. (Hint: you will need to look at Utilities/Variables to see how they are coded before interpreting your output)
       

   2. Did adding these variables improve the goodness of fit of the model? On what evidence do you base your model?
       

   3. What effect does being a women have on salary?
       

   4. What effect does race have on salary?
       

   5. Of the three independent variables, which has the greatest impact on salary? What number tells you this?
       

7. The agency HR director argues that the effect that gender and race have on salary is really due to women and minorities holding different (i.e., lower paid) jobs than men and nonminorities

   1. Is there a variable in this dataset that could be used to test this assumption? Can we include it in the model? Why or why not?
       

   2. Is there another means we could use to examine whether the difference between men/women's minorities/nonminorities is a function of occupational differences? If so, please do so and report your results.