POLS 6482 ADVANCED MULTIVARIATE STATISTICS
Ninth Assignment
Due 5 November 2001

  1. Download the EVIEWS dataset for the coffee example discussed in Epple Notes VI-20 to VI-24

    Coffee Data (coffee.wf1)

    and paste it into STATA. The variable cons is the per capita consumption of coffee in pounds, price is the price in cents per pound, pcinc is the per capita income in dollars, and year is the year.

    1. Turn in the d and summ commands.

    2. Replicate the regressions shown on pages VI-20, VI-22, and the Ramsey Reset tests shown on page VI-23.

    3. Interpret the coefficients for the Log-Log model. Do the signs on the coefficients make sense? Why? Why not?

  2. In this problem we will work with the Drinking Age and Highway Fatality rate dataset discussed in Epple Notes X-4 to X-16.

    Drinking Age and Highway Fatality rate dataset (EVIEWS Dataset)

    Download the dataset and bring it up in EVIEWS.

    1. Reproduce the results shown on page X-5 to X-15. Specifically, run the regression, do the White test, run the regression with the White Standard Error Correction, and do the weighted regression.

    2. Do the estimated coefficients make sense to you? Why or why not?

    3. Paste the dataset into Stata, define the variables appropriately, and turn in the d and summ commands.

    4. In Stata run the regressions shown on pages X-5 and X-11. To do the standard error correction use the command:

      regress lft18t20 tax drkage pcinc miles yngdrv insp mormon prot cath sobab wet, robust

    5. Produce the plot shown in Epple Notes X-6. In Stata you can do it with the commands:

      predict yresid, residuals

      plot yresid tax

      The first command places the residuals into the vector yresid and the second command produces a scatterplot with tax as the horizontal axis with the residuals on the vertical axis (note that this is backwards from EVIEWS!).

  3. In this problem we are going use the 1968 and 1996 NES presidential election data from homework 2 to test whether or not the same linear model applies to the party identification of men and women. Recall that the specification was:

    Party = f(income, race, sex, south, education, age)

    or, expressed in terms of a regression equation:

    y = b0 + b1x1 + b2x2 + b3x3 + b4x4 + b5x5 + b6x6 + e

    where y = party, x1 = income, x2 = race, x3 = sex, x4 = south, x5 = education, and x6 = age.

    Use the method shown in Epple Notes VII-2 to VII-8 to test whether or not women and men have different linear models for party identification. The hypothesis test is the same as that shown on page VII-7. In this context the indicator variable sex plays the same role as the indicator variable DUM in the delivery dataset.

    1. Do the hypothesis test in both EVIEWS and Stata for the 1968 and 1996 data. Calculate the exact p-value of the test using @fdist(f_stat, df_numerator, df_denominator) in EVIEWS and display fprob(df_numerator, df_denominator, f_stat) in Stata. Show all of your regression output and calculations.

    2. Discuss the substantive significance of these test results. Be specific.