Multiple Regression

Please answer the questions directly on this sheet, and attach your output.

Open Employee.sav in the PA 710 folder

 After looking at previous research as to the determinants of salary levels in organizations, we form an initial hypothesis.  Our hypothesis is that salary is at least partly a function of the years of education an employee has.  Test this hypothesis with a scatterplot.

1. What does the scatterplot suggest is the relationship (if any) between salary and education?
2. Run a bivariate regression procedure to test the hypothesis.

a)      Is our hypothesis supported?

b)      What would your best guess for salary be if the employee has no education?

c)      For each year of education, how much does salary increase?

3 . Suppose after doing some additional research, we find that salary is often also dependent on one’s seniority in the organization.  Add this second independent variable (months since hire). How much, if any, does seniority contribute to the explanatory power of our model?

4. Let’s suppose we found out that months since hire had not been computed correctly.  Take it out of the model.  Then suppose we also did some additional research and we found that gender and race often also affect salaries in an organization.  (Hint: check Utilities/Variables and make a note of how these variables are coded)

a)      State a hypothesis for what you would expect the relationship (if any) to be between these two variables and salary.

b)      Run the regression model again, adding gender and race as independent variables.  Were your hypotheses supported?

c)      Did adding these variables improve the goodness of fit of the model?  What evidence is there to support your answer?

d)      What effect does being a woman have on one’s salary?

e)      What can you say about the effect of race on one’s salary?

f)        Of the three independent variables, which has the greatest impact on salary?  What number(s) tell you this?

g)      Could we have used GENRACE to test this hypothesis? (Check Utilities/Variables to see how it is coded)  If not, why not

5. It has been suggested that the effect for race and gender on salary is really due to women and minorities holding different (lower paid) jobs in the organization than men and nonminorities.

a)      Is there a variable in this dataset that could be used to test this hypothesis?

b)      Can we simply include it in the regression model?  If not, why not?

c)      Is there some other means could we use to examine whether the difference in salary between men/women or minorities/nonminorities is caused by occupational differences?  If so, please do so and explain your results.