Multiple Regression Analysis: Florida Presidential Election and SAT Scores | Assignments Political Science

PSC 104 (10): Methods of Public Policy Analysis

Data Analysis Assignment 3

Due: December 6, 2002

(I) Florida 2000 data

Below are the results of a multiple regression using the 2000 presidential election returns for the

67 Florida counties. The number of Buchanan votes (Buchanan) is regressed on the total number

of votes (totalvote), the total number of Gore votes (gorevotes) and the total number of Bush

votes (bushvotes). Interpret all of the coefficients, both in terms of substantive effect and

statistical significance, given alternative hypotheses of the slopes not equaling zero.

. reg buchanan totalvot gore bush

Source | SS df MS Number of obs = 67

-------------+------------------------------ F( 3, 63) = 32.16

Model | 8062724.71 3 2687574.90 Prob > F = 0.0000

Residual | 5264801.92 63 83568.2844 R-squared = 0.6050

-------------+------------------------------ Adj R-squared = 0.5862

Total | 13327526.6 66 201932.222 Root MSE = 289.08

------------------------------------------------------------------------------

buchanan | Coef. Std. Err. t P>|t| [95% Conf. Interval]

-------------+----------------------------------------------------------------

totalvote | .1620454 .0357214 4.54 0.000 .0906619 .2334289

gorevotes | -.159742 .0361842 -4.41 0.000 -.2320503 -.0874337

bushvotes | -.1658197 .0365267 -4.54 0.000 -.2388125 -.0928269

_cons | 46.75452 46.27939 1.01 0.316 -45.72746 139.2365

------------------------------------------------------------------------------

Now we add a dummy variable for Palm Beach county (palmbeach). Again interpret all of the

coefficients, both in terms of substantive effect and statistical significance, given alternative

hypotheses of the slopes not equaling zero. Does this model fit the data better than the one

without the Palm Beach dummy? How do you know?

. reg buchanan totalvot gore bush palm

Source | SS df MS Number of obs = 67

-------------+------------------------------ F( 4, 62) = 467.51

Model | 12899840.2 4 3224960.06 Prob > F = 0.0000

Residual | 427686.389 62 6898.16757 R-squared = 0.9679

-------------+------------------------------ Adj R-squared = 0.9658

Total | 13327526.6 66 201932.222 Root MSE = 83.055

------------------------------------------------------------------------------

buchanan | Coef. Std. Err. t P>|t| [95% Conf. Interval]

-------------+----------------------------------------------------------------

totalvote | .0794995 .010726 7.41 0.000 .0580586 .1009404

gorevotes | -.0801572 .0108217 -7.41 0.000 -.1017894 -.058525

bushvotes | -.078031 .0110056 -7.09 0.000 -.1000308 -.0560311

palmbeach | 2596.966 98.07085 26.48 0.000 2400.925 2793.007

_cons | 49.58643 13.29682 3.73 0.000 23.00646 76.16639

------------------------------------------------------------------------------

Multiple Regression Analysis: Florida Presidential Election and SAT Scores, Assignments of Political Science

Related documents

Partial preview of the text

Download Multiple Regression Analysis: Florida Presidential Election and SAT Scores and more Assignments Political Science in PDF only on Docsity!

PSC 104 (10): Methods of Public Policy Analysis

Data Analysis Assignment 3

Due: December 6, 2002

(I) Florida 2000 data

Below are the results of a multiple regression using the 2000 presidential election returns for the

67 Florida counties. The number of Buchanan votes (Buchanan) is regressed on the total number

of votes (totalvote), the total number of Gore votes (gorevotes) and the total number of Bush

votes (bushvotes). Interpret all of the coefficients, both in terms of substantive effect and

statistical significance, given alternative hypotheses of the slopes not equaling zero.

Now we add a dummy variable for Palm Beach county (palmbeach). Again interpret all of the

coefficients, both in terms of substantive effect and statistical significance, given alternative

hypotheses of the slopes not equaling zero. Does this model fit the data better than the one

without the Palm Beach dummy? How do you know?

(II) SAT data

(a) Using data from the 50 states, you regress the average SAT score for each state

(TotalSAT) on per pupil expenditures, in thousands (perpupexp). Assume that you did this to

test whether money spent on education affects SAT scores. What do you conclude, using a.

level of significance? What is the substantive effect of per pupil expenditures on SAT scores?

(b) Now you add the percentage of students in the state taking the SAT test (pcttakeSAT).

Are the independent variables significant at the .05 level? How do you know? What is the

substantive impact of per pupil expenditures? Do you obtain different results than in part (a)? If

so, what might explain the discrepancy? If not, why are the results the same?

(b) Now assume that theory suggests that this relationship may vary by gender. To see

whether that is the case, you perform a three way cross-tabulation, now controlling for

gender. What are the results of the chi-squared tests for males and females? [You need to

perform two separate tests here.] Do you reach a different conclusion than you did in

part (a)?

. by gen: tab NY region, col chi

(c) Do you think the survey is a nationally representative survey? Why or why not?

(IV) Multivariate Analysis [data: your data from analysis 2, with one more variable.]

Take the dependent variable and independent variable from the data that you collected

and add another independent variable to the analysis. Does adding the new variable

change the conclusion you drew in the last assignment? Why or why not? Is there a

statistically significant relationship between each of your independent variables and your

dependent variable? Why or why not?