PUAF 610

Quantitative Methods in Policy Analysis

Fall 2007

Problem Set #10

Using the Milwaukee.xls data set, reproduce the results discussed in class for the effect of the Choice program on reading and math scores. You should do two separate regressions: one for reading and one for math.

A.

For each regression, discuss the effect of the Choice program on test score. What is your overall conclusion about the effect of the program?

B.

For one of the regressions, interpret the meaning of the other regression coefficients (i.e., for the control variables). Are the values what you expect?

C.

D.

For the same regression, plot and analyze the residuals.

Construct linear regression models using the UMCPsalaries.xls data set, with salary as the response variable.

A.

Carefully examine the definitions of the variables in the codebook. Which variables are redundant (remembering that all the dummy variables of a categorical variable cannot be included)? (If you are using Data Analysis, delete these variables from the data set or move them to one side.)

B.

Construct a correlation matrix with the remaining variables (except the department variables). Which explanatory variables are most strongly correlated with salary? Which explanatory variables are most correlated with each other?

C.

Construct a regression model using the following explanatory variables: yrdeg, male, black, asian, hisp, citizen, full, assoc, and phd. For which of these variables would you reject the null hypothesis (i.e., no correlation with salary)? Explain the meaning of these coefficients in plain English. Do these values make sense? Interpret R2 and the standard error in plain English.

D.

Construct a regression model that includes the above variables and the dummy variables for the colleges. Interpret the regression coefficients. What do you find?

NOTE: Data Analysis/Regression is limited to 16 explanatory variables; here we have 21. For our purposes here, just include as many as you can.

E.

Analyse the residuals. What do you conclude? What could be done to improve the regression?

F.

Use the regression equation in part D to predict Dr. Fetter's salary in Fall 1997, when he was a 37-year-old associate professor. He received his PhD in 1985.