Assignment 6 . Multiple Regression – Part II

1. Using SPSS and data from the 1992 Health and Retirement Study, select respondents for your analysis who are currently working for pay (i.e., use a filter as you did in assignment 5).
2. Select cases with complete data on the following variables: income, marital status, sex, race, age, education, and experience (i.e., use the ‘select if not missing’ syntax).
3. Create dummy variables for marital status (create one for each category of marital status), sex, and race.
4. Grand mean center age, education, and experience.
5. Create interaction effects for sex*race and experience*race.
   1. Sex and race interactions – you should have four sex*race interactions when you are done (e.g., female*white, female*black, female*Hispanic, and female*other).
   2. Experience and race interactions – Use the centered experience variable and the four race dummy variables to create the experience*race interactions – you should have four experience*race interactions when you are done (e.g., experience*white, etc.).
6. Create a squared term for age (using the centered age variable).
7. Examine the relationships between income (the dependent variable) and marital status, sex, race, age, education, and experience (the independent variables).

Interactions:

You expect that the relationships between sex and income and between experience and income differ by race – that is, you believe that the sex gap in earnings differs by race and that the ‘return’ for experience differs by racial group.

Non-linear relationships:

You also expect a non-linear relationship between age and income – previous research suggests that younger and older workers earn less than middle-aged workers.

Controls:

Don’t forget to control for marital status (treat married as the reference category) and education (which should be centered).

Run one multiple regression model in SPSS. Be careful when entering the various dummy variables, interaction effects, and squared terms!

Hint: your model should include a total of 18 variables (i.e., 11 main variables, 6 interaction variables, and 1 squared variable).

1. The y-intercept (you do not need to test for significance from 0).
2. Sex and race – Which, if any, of the interactions are statistically significant? If none are significant, what does this suggest about the relationships between sex and income and between race and income? I nterpret only the significant interactions . Treat sex as the ‘variable of interest’ and discuss how the relationship between sex and income differs by race (although it would be equally valid to discuss how the relationship between race and income differs by sex) – for example, for which category of race is the sex gap in income largest? You might find it useful to calculate: 1) the sex slope for White, Black, Hispanic, and other respondents and 2) the predicted income for all possible sex*race combinations. Once these (especially #2) have been calculated, you can easily calculate the sex gap in earnings separately for each category of race.
3. Experience and race – Which, if any, of the interactions are statistically significant? If none are significant, what does this suggest about the relationships between experience and income and between race and income? I nterpret only the significant interactions. Treat experience as the ‘variable of interest’ and discuss how the relationship between experience and income differs by race (although it would be equally valid to discuss how the relationship between race and income differs by experience). You might find it useful to calculate the experience-income slope for all categories of race.
4. Age – Is the squared variable statistically significant? If it is, interpret the non-linear relationship. How would you describe the relationship (e.g., is it u-shaped, etc.)? What is the point of inflection – that is, at what age does the relationship change direction? What is the ‘slope’ when age is two standard deviations below the mean, at the mean, and two standard deviations above the mean?
5. Marital status – Are any of the differences statistically significant? If so, interpret them.
6. Education – Is the relationship statistically significant? If it is, interpret the linear relationship.
7. The global F test (#3 in the handout from Topic 9) – Identify the null and research hypotheses for the global F test and make and explain your decision (i.e., should the null hypothesis be rejected or not).
8. Adjusted r-squared – how much variation in income does your model explain?
9. How might you trim this model, if at all? Note: You do not need to estimate an additional trimmed model. I just want you to suggest a change based on your results.

Note: please send your syntax file to me via e-mail. You do not have to print a hard copy!!!

Price: $24.0

Solution: The downloadable solution consists of 10 pages, 1400 words and 4 charts.
Deliverable: Word Document